`

Noise-aware Unsupervised Anomalous Sound Detection for Machine Condition Monitoring

Challenge results

Task description

The aim of this task is to develop noise-robust unsupervised anomalous sound detection systems by exploiting recordings captured simultaneously by multiple microphones placed at different distances from the target machine. More detailed task description can be found in the task description page

Teams ranking

Table including only the best performing system per submitting team.

Rank Submission Information Evaluation Dataset Development Dataset
Submission Code Technical
Report 		Official
Rank 		Official
Score 		ToyDrone
(AUC) 		ToyDrone
(pAUC) 		ToothBrush
(AUC) 		ToothBrush
(pAUC) 		SewingMachine
(AUC) 		SewingMachine
(pAUC) 		Sander
(AUC) 		Sander
(pAUC) 		BlowerDustCollector
(AUC) 		BlowerDustCollector
(pAUC) 		fan
(AUC) 		fan
(pAUC) 		gearboxEmu
(AUC) 		gearboxEmu
(pAUC) 		bearingEmu
(AUC) 		bearingEmu
(pAUC) 		sliderEmu
(AUC) 		sliderEmu
(pAUC) 		ToyCarEmu
(AUC) 		ToyCarEmu
(pAUC) 		ToyCar
(AUC) 		ToyCar
(pAUC) 		valveEmu
(AUC) 		valveEmu
(pAUC)
DCASE2026_baseline_task2_MSE DCASE2026baseline2026 68 59.80289616592438 ± 0.0026619662647486666 56.50 55.11 61.57 59.79 63.08 58.11 53.18 50.89 77.80 61.79 54.20 53.33 59.01 52.94 60.95 59.85 56.15 50.38 65.41 55.89 56.75 54.03 68.26 55.08
Ozeki_MELCO_task2_4 OzekiMELCO2026 49 60.968307314170644 ± 0.0027324691850408502 57.98 55.21 58.76 54.84 65.34 57.42 64.05 53.11 77.34 58.11 49.06 50.74 66.44 51.53 57.46 55.21 60.31 52.79 54.74 50.32 67.38 52.05 73.42 56.37
Qian_SJTU_task2_2 QianSJTU2026 54 60.685217554546256 ± 0.002868576428036427 58.92 53.68 58.08 52.37 59.33 57.89 62.70 54.58 78.85 68.95 72.70 61.74 76.62 58.47 59.16 53.95 54.95 51.11 67.81 56.05 71.03 54.16 73.43 61.11
Qian_VUILabs_task2_1 QianVUILabs2026 55 60.61215647721782 ± 0.0027688037966323107 65.80 59.84 54.03 50.32 55.17 54.47 63.08 50.95 85.24 71.05 78.47 61.63 74.93 62.05 60.97 54.42 59.78 51.84 71.51 49.21 67.04 57.95 66.74 50.47
Zhang_XJTLU_task2_4 ZhangXJTLU2026 131 55.17563151568782 ± 0.002798601551388066 54.10 53.11 54.53 49.21 61.15 55.84 54.90 51.37 58.65 54.21 56.56 56.32 65.02 56.21 55.00 56.05 55.94 53.05 68.00 57.16 66.40 56.00 87.70 75.42
Chang_Surrey_task2_1 ChangSurrey2026 58 60.32203484103438 ± 0.0026443592346275958 43.66 52.63 61.19 56.11 71.70 55.42 64.19 56.95 80.68 71.89 75.14 52.74 68.80 54.74 65.80 56.21 63.74 52.26 78.58 63.42 82.82 61.37 81.06 59.47
Zhang_SATLab_task2_3 ZhangSATLab2026 39 61.24923201074941 ± 0.002952301148904951 73.48 60.95 58.59 52.79 49.73 53.00 65.69 56.37 79.71 63.16 73.12 59.58 80.94 71.42 64.78 56.79 61.48 50.95 87.14 67.79 76.55 62.79 74.01 54.11
Fan_WISTLAB_task2_4 FanWISTLAB2026 17 62.99413985866293 ± 0.003148125130529185 68.27 61.26 59.34 53.63 56.73 57.42 65.15 53.84 83.19 70.58 68.79 60.42 80.68 64.32 65.34 60.32 66.39 51.32 72.03 53.32 81.44 67.11 80.08 56.63
Jiang_AITHU_task2_3 JiangAITHU2026 21 62.3631416226516 ± 0.00304945479733028 72.59 62.16 57.97 52.95 55.67 56.21 63.77 53.47 79.73 68.32 71.80 60.05 80.72 67.74 65.50 59.68 63.48 51.74 81.15 56.58 79.71 67.37 78.93 57.53
Zhang_THUEE_task2_4 ZhangTHUEE2026 11 63.71007071087524 ± 0.003109602589321203 69.98 62.37 59.67 54.89 59.95 60.05 63.63 52.79 83.14 68.26 69.74 58.11 79.37 62.79 65.89 61.21 65.37 52.00 76.63 53.58 81.13 70.84 79.34 58.05
Huang_CQUPT_task2_4 HuangCQUPT2026 109 57.345977001988956 ± 0.003004740948360317 58.20 56.53 45.14 49.95 65.05 62.63 54.71 50.68 72.46 68.53 57.43 51.26 61.56 54.32 59.37 58.68 69.05 55.58 66.48 50.05 78.24 57.95 83.99 74.84
Xie_SHU_task2_2 XieSHU2026 65 60.03523389769877 ± 0.002807957195771089 51.20 51.26 56.77 51.89 69.51 61.68 64.30 61.84 73.18 58.16 53.00 53.53 65.46 52.21 62.49 55.00 72.02 59.79 69.74 57.05 74.60 60.11 83.13 70.89
Moon_Independent_task2_1 MoonIndependent2026 127 55.60880820594806 ± 0.002826442080110754 54.68 55.00 52.37 54.26 62.32 60.00 56.04 50.79 58.54 49.84 64.66 50.58 70.72 64.00 63.12 60.42 62.20 53.42 79.26 57.37 70.14 54.58 90.16 83.21
Xia_NEU_task2_1 XiaNEU2026 32 61.72131772891849 ± 0.002899363535449049 63.93 53.84 50.03 50.68 68.34 69.16 58.25 51.47 86.85 77.21 63.90 53.43 74.26 57.45 56.90 49.78 75.68 55.60 81.18 50.22 73.24 57.29 92.46 84.75
XingWu_MCPX_task2_4 XingWuMCPX2026 29 61.87465528410395 ± 0.002766292907182805 63.29 57.84 57.38 51.37 65.61 61.47 59.63 51.84 79.51 69.05 67.08 54.11 68.08 59.11 58.74 54.58 56.18 48.68 70.26 52.53 77.86 56.53 67.12 52.11
Zhou_SUMERUZOO_task2_3 ZhouSUMERUZOO2026 174 48.43662010700282 ± 0.002323678592788652 40.43 51.00 48.91 49.11 49.73 47.58 50.01 50.11 51.60 51.53 68.36 55.53 52.99 49.32 63.63 53.63 61.00 55.16 58.00 56.53 56.01 52.79 69.01 61.05
Kwon_KIST_task2_1 KwonKIST2026 110 57.33672495384062 ± 0.0028127046901336617 46.02 50.11 52.08 52.89 57.60 62.00 61.68 53.05 80.05 72.58 69.70 61.42 72.10 61.11 68.40 61.32 67.18 56.74 75.10 52.47 80.68 63.95 91.96 78.42
Fujimura_MERL_task2_3 FujimuraMERL2026 1 70.24115096031407 ± 0.0030780430693608625 71.79 63.53 68.36 61.74 71.45 68.79 69.52 57.37 88.64 75.74 57.58 51.21 77.45 58.00 67.15 62.05 75.33 69.42 70.59 51.05 65.32 52.00 98.20 91.89
Noh_CBNU_task2_1 NohCBNU2026 118 56.88833422457569 ± 0.002686961640274201 58.01 54.47 52.33 53.53 61.84 57.21 55.82 54.32 65.75 51.37 53.99 54.16 72.77 62.26 65.95 59.37 57.73 53.08 67.81 54.05 65.97 52.29 89.32 76.05
Jeong_KETI_task2_1 JeongKETI2026 130 55.25280412349333 ± 0.0027578476609583817 51.41 55.05 58.24 51.00 49.51 50.58 58.25 56.05 63.74 59.74 42.86 50.58 62.00 56.53 58.53 53.37 67.43 56.42 69.13 56.84 61.34 51.63 53.48 52.79
Zarrouky_IR_task2_1 ZarroukyIR2026 140 53.833412450926346 ± 0.0025414134115744453 59.02 51.68 48.80 50.13 54.33 51.34 48.17 48.37 70.14 57.26 91.98 73.05 52.73 50.53 60.54 57.08 60.06 50.97 57.28 51.84 65.42 52.79 76.06 54.00
Lei_CRRC_task2_4 LeiCRRC2026 108 57.38198513172906 ± 0.0028700109628889314 44.14 50.89 56.21 51.58 66.27 62.42 59.52 54.68 72.45 63.37 59.55 52.89 71.84 62.47 55.73 60.16 68.22 56.95 67.64 50.05 79.64 62.47 78.85 58.37
Tsz_HFU_task2_3 TszHFU2026 103 57.51642313488529 ± 0.0027109071470683395 61.22 50.79 60.00 53.26 60.70 55.58 57.79 53.84 61.52 50.79 44.62 50.11 68.47 55.00 53.67 52.42 56.34 49.16 67.67 52.05 64.57 49.42 72.72 55.95
Kim_LUDO_task2_3 KimLUDO2026 53 60.72626822655016 ± 0.002906834203180997 58.06 63.47 55.78 54.58 64.36 63.42 53.28 51.68 79.35 73.00 49.04 52.95 62.73 53.11 61.80 59.95 71.00 55.37 70.48 49.74 81.21 60.26 81.09 69.00
Balozi_RISE_task2_2 BaloziRISE2026 154 51.297542290502165 ± 0.0025676312077536315 34.96 49.74 52.06 48.95 60.60 60.05 53.25 50.68 64.55 54.58 49.39 49.89 63.11 56.95 52.97 52.11 57.88 53.89 61.62 51.95 71.16 58.84 75.97 55.32
Mei_FDID_task2_1 MeiFDID2026 129 55.39000249761615 ± 0.0026175082253723754 55.07 50.68 49.14 50.53 66.13 58.89 52.74 49.53 63.83 56.84 81.75 61.40 60.10 54.00 62.40 60.00 59.60 50.70 75.00 58.20 75.25 55.10 85.90 61.70
Wang_WST_task2_3 WangWST2026 145 53.13694610777225 ± 0.00274928682418755 49.31 53.05 46.28 50.84 53.56 53.16 48.39 49.68 74.71 64.05 63.20 52.42 66.40 56.32 63.74 61.63 56.06 50.11 73.26 57.58 70.96 54.42 67.72 53.74
Jiang_KY_task2_3 JiangKY2026 141 53.558263051309616 ± 0.002623267856524193 53.13 52.26 51.86 48.89 48.87 51.74 56.61 50.21 66.78 53.11 70.56 54.47 73.12 60.79 56.52 49.32 50.20 49.84 41.66 47.68 45.88 50.42 50.46 51.11
Glitza_IKA_task2_3 GlitzaIKA2026 139 54.4613765901715 ± 0.0025487340237951833 54.50 50.63 47.73 51.53 56.92 53.71 56.39 50.26 67.43 52.95 57.42 51.68 58.36 49.37 60.66 59.16 58.20 48.79 56.10 53.47 59.53 51.00 44.24 49.34
Kajita_IND_task2_3 KajitaIND2026 151 51.77972417170499 ± 0.00243164973546122 60.64 53.58 48.93 49.76 53.20 51.53 52.35 50.11 47.61 50.26 56.38 50.58 53.32 51.26 53.13 50.21 47.08 49.89 56.05 50.58 45.83 48.37 48.01 49.95
Zhang_JAIST_task2_1 ZhangJAIST2026 47 61.02577712684993 ± 0.0030562402643252644 57.52 54.42 60.15 51.00 67.28 64.37 57.56 53.47 75.21 69.00 64.61 54.74 70.55 62.58 64.56 60.42 70.90 58.58 69.07 53.63 79.00 58.79 79.77 64.53
Wang_UniS_task2_1 WangUniS2026 72 59.593310968513094 ± 0.0027704226868877947 59.58 56.68 53.05 49.05 58.77 57.63 55.90 52.16 86.01 78.00 63.36 52.42 70.32 53.21 64.00 60.00 63.56 49.16 78.04 60.00 76.10 61.84 81.16 68.26
Yang_XJU_task2_4 YangXJU2026 62 60.15236729958677 ± 0.002637242776897266 55.01 52.21 51.72 48.21 63.68 62.95 63.78 52.68 86.28 73.37 59.47 51.47 76.32 59.26 62.53 60.32 62.90 50.79 77.64 62.21 79.94 66.58 79.65 73.79
SNU_task2_1 SNUtask22026 169 50.02291439923236 ± 0.0026525044143756536 40.88 51.79 48.09 48.95 44.94 49.79 55.97 53.42 60.84 56.53 61.56 54.37 65.14 55.58 61.66 51.47 61.18 53.47 74.20 59.16 77.92 50.53 83.86 63.32
Huang_WHU_task2_1 HuangWHU2026 4 65.45291026592781 ± 0.003042006671244826 69.29 58.89 64.22 55.16 65.60 65.42 63.84 55.63 80.24 71.16 74.08 53.89 78.21 54.58 74.85 64.95 61.99 52.79 73.48 52.21 80.07 61.11 79.35 52.84
Morita_KM_task2_4 MoritaKM2026 90 58.11117958443154 ± 0.0027940480488639475 54.92 51.74 57.97 52.58 61.25 51.84 57.26 50.58 75.11 64.16 78.03 55.00 65.11 56.21 62.58 60.68 61.91 52.58 68.84 58.53 62.37 56.63 74.01 53.37
Wu_CUMT_task2_4 WuCUMT2026 3 65.46214466728418 ± 0.0032688205174339403 64.17 57.05 63.74 56.16 69.51 59.05 63.16 55.89 91.32 70.95 78.88 60.53 79.55 65.68 61.89 53.79 59.49 49.63 79.75 60.68 76.71 62.05 74.89 58.16
Yang_None_task2_1 YangNone2026 10 64.29601690030654 ± 0.0031399217014158464 65.01 56.26 58.39 50.95 68.50 58.84 62.76 55.68 90.32 78.63 61.06 60.58 79.18 64.21 58.57 51.74 54.65 49.21 79.17 61.53 73.33 63.11 78.52 67.95
Zheng_HFUUAI_task2_3 ZhengHFUUAI2026 25 62.113376977891996 ± 0.0028591933097166624 68.39 58.84 62.36 51.63 61.53 56.05 59.83 52.68 76.98 66.32 67.51 50.84 72.31 54.79 60.58 60.11 55.14 48.84 65.76 55.00 74.88 62.47 84.76 61.95
Zhou_HFUUDS_task2_4 ZhouHFUUDS2026 20 62.4013149311637 ± 0.0028683405358712537 67.63 57.37 60.31 52.05 61.95 56.74 60.92 53.74 79.98 68.26 64.94 51.79 71.97 55.84 58.94 60.68 56.70 49.26 65.85 54.26 76.21 62.00 86.54 67.11
Zhu_FDA_task2_4 ZhuFDA2026 94 57.81527446524864 ± 0.0028700545499872093 50.49 50.32 51.30 51.74 62.09 59.89 59.72 54.16 74.40 71.95 64.33 52.37 65.33 57.63 65.80 57.47 67.30 52.42 63.73 57.68 60.34 54.05 81.01 67.16
Huang_QWS_task2_1 HuangQWS2026 27 62.0432723795996 ± 0.002884666717481283 59.84 53.74 56.96 49.68 65.46 64.05 61.25 55.26 85.98 69.84 64.38 58.00 72.19 63.32 66.10 59.05 64.69 53.58 63.22 55.53 69.28 55.61 92.86 80.63
Wang_Liu_SuzhouDongyuan_task2_1 WangLiu2026 146 53.07523562924925 ± 0.002651215737798937 50.61 51.84 44.91 49.42 55.24 55.32 55.86 50.37 65.33 54.11 64.36 52.37 67.52 53.21 57.60 55.47 52.68 48.79 69.82 55.26 77.26 63.68 79.08 59.37
Qian_nivic_task2_2 Qiannivic2026 8 64.42767291560604 ± 0.002963997006229839 72.85 62.47 49.35 50.26 69.47 58.42 66.37 59.26 87.24 76.37 73.41 58.84 77.56 60.63 56.52 52.16 62.42 50.26 73.65 57.05 74.89 57.37 76.39 60.73
Jeong_Medisensing_task2_4 JeongMedisensing2026 106 57.435901782845434 ± 0.0029330489767252477 52.95 53.11 51.79 50.42 57.55 55.11 54.39 50.61 84.19 77.37 65.62 53.05 67.91 57.74 60.54 58.16 60.54 51.53 69.89 51.21 69.31 56.24 84.93 57.26
Zhou_XAUAT_task2_1 ZhouXAUAT2026 37 61.37081867585763 ± 0.0027780323220549705 65.09 56.58 53.72 52.42 58.92 54.58 66.22 57.95 80.16 66.53 85.83 74.74 89.71 81.58 63.69 59.63 63.95 52.74 70.44 54.74 66.58 55.37 61.87 52.63
Kim_CAU_task2_2 KimCAU2026 66 59.85423786944168 ± 0.003219819206690976 69.73 56.21 49.85 48.26 57.25 61.21 59.71 54.21 84.27 61.16 62.00 57.00 66.12 53.16 59.46 54.47 62.82 51.16 78.00 60.11 72.98 60.84 79.56 62.53
Zeng_BUCT_task2_4 ZengBUCT2026 162 51.10928407930502 ± 0.0024914923213899603 48.80 52.00 45.74 47.95 56.52 51.79 48.65 52.37 58.62 50.84 56.26 53.63 74.60 62.32 53.92 54.21 59.38 50.74 70.04 62.42 75.40 58.26 84.72 77.79
Yang_NJU_task2_1 YangNJU2026 43 61.1737276780872 ± 0.002856466985997388 71.07 57.11 45.68 48.16 57.26 51.16 72.62 60.84 85.05 76.74 0.60 0.58 0.69 0.55 0.62 0.56 0.59 0.50 0.73 0.59 0.82 0.63 0.67 0.52
Kim_KATECH_task2_3 KimKATECH2026 61 60.21000436812732 ± 0.0023850928221648194 60.92 59.58 53.44 56.74 61.44 54.42 57.83 50.58 82.26 67.26 75.70 54.68 71.15 58.32 62.39 59.32 61.28 51.95 67.35 50.47 78.37 60.58 53.37 50.37
Guan_GISP@HEU_task2_1 GuanGISP@HEU2026 79 59.00784814563139 ± 0.0031432640274972243 60.59 61.11 55.98 53.53 62.48 59.11 49.21 48.74 77.71 68.53 60.50 54.20 60.25 56.10 60.50 58.90 59.65 50.00 71.90 51.30 77.40 56.10 70.95 50.60
Krag_AAU_task2_4 KragAAU2026 19 62.58781647688 ± 0.0029384679344328225 64.44 58.24 62.35 54.61 58.76 54.13 61.76 54.50 81.92 74.58 67.34 52.53 79.66 68.34 64.63 57.84 70.91 55.45 80.91 71.13 80.38 63.21 89.61 81.45
Moradi_JKU_task2_1 MoradiJKU2026 52 60.79515879820278 ± 0.003065461869000823 57.63 53.21 52.47 53.95 70.82 65.53 66.08 53.53 72.00 61.16 59.68 58.58 63.02 52.89 63.36 62.21 62.84 54.74 70.02 49.11 70.62 53.47 81.30 72.42


Supplementary metrics (recall, precision, and F1 score)

Rank Submission Information Evaluation Dataset
Submission Code Technical
Report 		Official
Rank 		ToyDrone
(F1 score) 		ToyDrone
(Recall) 		ToyDrone
(Precision) 		ToothBrush
(F1 score) 		ToothBrush
(Recall) 		ToothBrush
(Precision) 		SewingMachine
(F1 score) 		SewingMachine
(Recall) 		SewingMachine
(Precision) 		Sander
(F1 score) 		Sander
(Recall) 		Sander
(Precision) 		BlowerDustCollector
(F1 score) 		BlowerDustCollector
(Recall) 		BlowerDustCollector
(Precision)
DCASE2026_baseline_task2_MSE DCASE2026baseline2026 68 58.30 61.61 55.32 66.67 100.00 50.00 61.68 72.66 53.59 65.73 86.44 53.03 69.94 98.99 54.07
Ozeki_MELCO_task2_4 OzekiMELCO2026 49 64.47 93.62 49.16 56.87 60.59 53.58 64.78 81.95 53.55 65.89 79.95 56.04 74.71 89.82 63.96
Qian_SJTU_task2_2 QianSJTU2026 54 56.00 56.00 56.00 58.00 58.00 58.00 54.96 54.84 55.08 56.56 56.56 56.56 73.74 73.92 73.57
Qian_VUILabs_task2_1 QianVUILabs2026 55 62.00 62.00 62.00 54.75 54.84 54.66 52.98 52.83 53.13 63.91 63.75 64.07 78.54 78.91 78.16
Zhang_XJTLU_task2_4 ZhangXJTLU2026 131 0.00 0.00 0.00 0.00 0.00 0.00 5.80 3.00 85.71 0.00 0.00 0.00 6.33 3.33 62.50
Chang_Surrey_task2_1 ChangSurrey2026 58 28.17 18.11 63.32 54.84 51.53 58.62 70.10 69.94 70.26 60.13 59.73 60.54 71.82 72.05 71.59
Zhang_SATLab_task2_3 ZhangSATLab2026 39 65.80 65.94 65.66 55.30 54.88 55.74 50.00 49.92 50.08 60.82 60.85 60.79 73.93 74.35 73.52
Fan_WISTLAB_task2_4 FanWISTLAB2026 17 62.55 62.60 62.50 51.25 50.72 51.79 53.78 53.70 53.86 61.98 61.94 62.02 76.79 77.18 76.40
Jiang_AITHU_task2_3 JiangAITHU2026 21 64.52 64.62 64.42 55.09 54.86 55.33 56.00 56.00 56.00 58.91 58.85 58.97 71.92 72.33 71.51
Zhang_THUEE_task2_4 ZhangTHUEE2026 11 61.59 60.32 62.91 57.92 57.38 58.47 56.95 56.98 56.92 59.06 58.85 59.27 74.92 75.16 74.69
Huang_CQUPT_task2_4 HuangCQUPT2026 109 65.32 96.99 49.24 65.98 97.96 49.74 63.43 79.95 52.56 64.73 83.24 52.95 67.35 98.00 51.31
Xie_SHU_task2_2 XieSHU2026 65 44.02 37.74 52.80 57.20 57.38 57.03 59.49 50.51 72.36 53.61 47.67 61.24 71.29 75.95 67.16
Moon_Independent_task2_1 MoonIndependent2026 127 52.19 48.86 56.02 47.69 44.68 51.14 56.72 57.63 55.85 59.03 58.93 59.13 55.85 56.90 54.84
Xia_NEU_task2_1 XiaNEU2026 32 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
XingWu_MCPX_task2_4 XingWuMCPX2026 29 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Zhou_SUMERUZOO_task2_3 ZhouSUMERUZOO2026 174 7.25 3.84 64.86 5.92 3.33 26.32 4.71 2.67 20.00 10.08 5.71 42.55 27.69 18.95 51.43
Kwon_KIST_task2_1 KwonKIST2026 110 47.25 40.92 55.88 52.15 52.15 52.15 52.00 52.00 52.00 60.00 60.00 60.00 68.54 68.64 68.44
Fujimura_MERL_task2_3 FujimuraMERL2026 1 57.53 47.36 73.26 64.98 71.74 59.39 59.89 51.69 71.19 56.27 46.47 71.33 75.47 100.00 60.61
Noh_CBNU_task2_1 NohCBNU2026 118 55.15 52.41 58.19 52.62 51.33 53.97 56.28 55.93 56.64 56.99 56.84 57.14 61.94 61.94 61.94
Jeong_KETI_task2_1 JeongKETI2026 130 28.04 21.12 41.71 55.62 53.33 58.11 17.09 10.00 58.82 42.44 30.97 67.42 38.57 31.11 50.72
Zarrouky_IR_task2_1 ZarroukyIR2026 140 54.03 53.93 54.13 49.20 48.82 49.59 54.15 53.53 54.78 50.17 50.51 49.83 64.39 63.75 65.03
Lei_CRRC_task2_4 LeiCRRC2026 108 66.67 93.62 51.76 65.86 80.99 55.50 61.05 65.79 56.95 62.99 74.88 54.36 65.10 88.90 51.36
Tsz_HFU_task2_3 TszHFU2026 103 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 40.21 28.97 65.73 31.29 20.95 61.80
Kim_LUDO_task2_3 KimLUDO2026 53 50.55 45.71 56.54 50.55 54.10 47.43 57.01 46.98 72.49 46.45 38.77 57.93 68.63 83.24 58.38
Balozi_RISE_task2_2 BaloziRISE2026 154 18.29 11.25 48.91 52.99 52.08 53.93 54.01 53.93 54.09 52.63 52.53 52.73 62.98 62.98 62.98
Mei_FDID_task2_1 MeiFDID2026 129 53.23 52.15 54.36 52.77 52.08 53.49 58.52 58.17 58.87 53.77 52.81 54.76 63.03 62.98 63.08
Wang_WST_task2_3 WangWST2026 145 7.15 3.73 84.85 20.00 12.00 60.00 22.97 13.71 70.59 6.62 3.60 40.91 28.77 16.80 100.00
Jiang_KY_task2_3 JiangKY2026 141 54.78 53.53 56.10 50.56 50.51 50.61 47.67 47.67 47.67 60.55 58.33 62.95 59.76 58.23 61.37
Glitza_IKA_task2_3 GlitzaIKA2026 139 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Kajita_IND_task2_3 KajitaIND2026 151 0.00 0.00 0.00 7.55 4.00 66.67 5.69 3.00 54.55 0.00 0.00 0.00 0.00 0.00 0.00
Zhang_JAIST_task2_1 ZhangJAIST2026 47 12.96 7.00 87.50 19.66 11.67 62.50 32.21 19.20 100.00 20.92 12.31 69.57 26.09 15.00 100.00
Wang_UniS_task2_1 WangUniS2026 72 39.23 28.64 62.25 34.34 27.10 46.88 44.27 34.74 60.98 50.89 44.44 59.52 67.10 98.99 50.76
Yang_XJU_task2_4 YangXJU2026 62 43.12 37.53 50.66 54.88 51.93 58.18 60.05 59.93 60.17 58.61 57.63 59.63 78.58 78.68 78.48
SNU_task2_1 SNUtask22026 169 32.97 24.30 51.27 17.98 12.00 35.82 23.88 16.00 47.06 35.04 26.67 51.06 58.29 53.93 63.41
Huang_WHU_task2_1 HuangWHU2026 4 31.66 20.21 73.00 54.00 46.53 64.33 47.29 30.97 100.00 42.60 30.19 72.33 66.27 53.93 85.95
Morita_KM_task2_4 MoritaKM2026 90 50.94 47.08 55.49 43.17 35.00 56.30 49.41 39.60 65.67 44.36 36.92 55.56 68.17 65.88 70.62
Wu_CUMT_task2_4 WuCUMT2026 3 59.24 58.23 60.29 56.00 56.00 56.00 66.02 65.94 66.10 54.98 54.98 54.98 83.39 83.01 83.78
Yang_None_task2_1 YangNone2026 10 60.47 58.94 62.08 56.72 56.84 56.60 63.01 62.86 63.16 56.87 56.84 56.90 78.66 78.91 78.41
Zheng_HFUUAI_task2_3 ZhengHFUUAI2026 25 17.82 9.88 90.32 19.34 11.67 56.45 28.24 16.94 84.71 23.33 14.00 70.00 13.33 7.16 97.14
Zhou_HFUUDS_task2_4 ZhouHFUUDS2026 20 18.07 10.00 93.75 20.92 12.31 69.57 28.26 16.94 85.21 21.46 12.92 63.16 22.70 12.80 100.00
Zhu_FDA_task2_4 ZhuFDA2026 94 24.32 14.44 77.08 25.85 16.47 60.09 42.83 29.87 75.68 48.27 38.97 63.39 67.04 68.57 65.57
Huang_QWS_task2_1 HuangQWS2026 27 26.75 16.15 77.78 26.13 17.68 50.00 49.32 34.97 83.61 45.23 32.94 72.16 57.53 42.00 91.30
Wang_Liu_SuzhouDongyuan_task2_1 WangLiu2026 146 57.00 59.53 54.67 47.56 44.44 51.15 56.76 60.98 53.08 53.32 54.04 52.63 66.89 100.00 50.25
Qian_nivic_task2_2 Qiannivic2026 8 66.71 65.88 67.55 47.88 47.92 47.84 66.00 66.00 66.00 61.44 60.97 61.93 73.07 73.39 72.75
Jeong_Medisensing_task2_4 JeongMedisensing2026 106 13.55 7.41 79.37 21.51 13.33 55.56 34.64 23.04 69.73 30.84 20.57 61.54 50.22 33.53 100.00
Zhou_XAUAT_task2_1 ZhouXAUAT2026 37 59.83 59.93 59.73 56.85 57.63 56.09 53.18 51.93 54.50 60.88 60.39 61.38 74.11 73.95 74.27
Kim_CAU_task2_2 KimCAU2026 66 33.52 21.76 72.92 0.00 0.00 0.00 40.11 28.80 66.06 42.41 30.71 68.49 77.82 78.38 77.26
Zeng_BUCT_task2_4 ZengBUCT2026 162 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Yang_NJU_task2_1 YangNJU2026 43 35.69 23.44 74.78 10.21 6.00 34.29 0.00 0.00 0.00 7.69 4.00 100.00 82.66 95.83 72.67
Kim_KATECH_task2_3 KimKATECH2026 61 66.67 100.00 50.00 66.67 100.00 50.00 58.87 65.94 53.18 63.61 83.81 51.25 64.82 84.05 52.75
Guan_GISP@HEU_task2_1 GuanGISP@HEU2026 79 47.66 37.53 65.28 51.65 56.14 47.83 51.04 38.97 73.93 44.92 37.74 55.46 70.31 82.02 61.52
Krag_AAU_task2_4 KragAAU2026 19 61.56 63.77 59.49 61.86 70.00 55.41 60.66 70.99 52.95 62.39 71.94 55.07 74.29 78.75 70.31
Moradi_JKU_task2_1 MoradiJKU2026 52 66.44 95.96 50.81 64.66 92.73 49.63 68.55 88.72 55.86 69.10 92.90 55.00 67.13 95.96 51.63



Systems ranking

Rank Submission Information Evaluation Dataset Development Dataset
Submission Code Technical
Report 		Official
Rank 		Official
Score 		ToyDrone
(AUC) 		ToyDrone
(pAUC) 		ToothBrush
(AUC) 		ToothBrush
(pAUC) 		SewingMachine
(AUC) 		SewingMachine
(pAUC) 		Sander
(AUC) 		Sander
(pAUC) 		BlowerDustCollector
(AUC) 		BlowerDustCollector
(pAUC) 		fan
(AUC) 		fan
(pAUC) 		gearboxEmu
(AUC) 		gearboxEmu
(pAUC) 		bearingEmu
(AUC) 		bearingEmu
(pAUC) 		sliderEmu
(AUC) 		sliderEmu
(pAUC) 		ToyCarEmu
(AUC) 		ToyCarEmu
(pAUC) 		ToyCar
(AUC) 		ToyCar
(pAUC) 		valveEmu
(AUC) 		valveEmu
(pAUC)
DCASE2026_baseline_task2_MAHALA DCASE2026baseline2026 137 54.76277174340481 ± 0.0026911528940989785 62.18 59.26 41.56 52.79 50.73 50.58 54.80 49.47 76.20 63.89 52.55 52.29 63.61 53.97 64.10 60.42 57.77 50.36 68.06 53.47 65.22 58.25 56.55 50.20
DCASE2026_baseline_task2_MSE DCASE2026baseline2026 68 59.80289616592438 ± 0.0026619662647486666 56.50 55.11 61.57 59.79 63.08 58.11 53.18 50.89 77.80 61.79 54.20 53.33 59.01 52.94 60.95 59.85 56.15 50.38 65.41 55.89 56.75 54.03 68.26 55.08
Ozeki_MELCO_task2_1 OzekiMELCO2026 121 56.806486243225564 ± 0.0027004679256412696 56.51 56.47 54.03 52.89 69.05 57.79 59.82 56.63 53.12 49.95 63.52 54.47 66.55 57.05 51.08 55.31 60.96 53.11 58.73 52.16 59.20 54.53 74.01 61.05
Ozeki_MELCO_task2_2 OzekiMELCO2026 107 57.38857959349032 ± 0.002755826988570763 53.16 54.32 66.45 54.00 62.29 53.16 59.62 55.00 55.92 54.63 44.39 49.89 61.74 54.02 55.34 54.68 65.33 56.93 66.51 55.60 61.85 52.04 55.03 53.20
Ozeki_MELCO_task2_3 OzekiMELCO2026 59 60.21998041720763 ± 0.002722957556933986 63.87 60.05 52.60 55.47 57.79 52.95 61.86 50.53 82.02 66.37 64.10 51.16 68.78 56.32 67.67 59.63 60.68 51.47 65.34 51.95 70.67 59.32 78.14 58.37
Ozeki_MELCO_task2_4 OzekiMELCO2026 49 60.968307314170644 ± 0.0027324691850408502 57.98 55.21 58.76 54.84 65.34 57.42 64.05 53.11 77.34 58.11 49.06 50.74 66.44 51.53 57.46 55.21 60.31 52.79 54.74 50.32 67.38 52.05 73.42 56.37
Qian_SJTU_task2_1 QianSJTU2026 89 58.21935235088495 ± 0.002527073940910703 67.20 56.68 53.15 49.63 51.31 53.89 66.11 55.47 71.68 53.32 74.80 62.37 78.93 64.26 63.50 54.58 59.67 51.58 76.66 60.58 67.75 60.26 66.44 51.26
Qian_SJTU_task2_2 QianSJTU2026 54 60.685217554546256 ± 0.002868576428036427 58.92 53.68 58.08 52.37 59.33 57.89 62.70 54.58 78.85 68.95 72.70 61.74 76.62 58.47 59.16 53.95 54.95 51.11 67.81 56.05 71.03 54.16 73.43 61.11
Qian_SJTU_task2_3 QianSJTU2026 91 58.0578727586588 ± 0.002779400309258946 69.64 60.74 50.35 50.05 47.64 51.95 67.20 54.37 74.97 57.42 73.73 61.95 78.64 69.58 63.63 54.47 60.60 51.11 78.46 61.32 69.43 62.26 67.49 52.37
Qian_SJTU_task2_4 QianSJTU2026 88 58.40458833787484 ± 0.0027547347298413506 69.93 62.32 49.20 49.47 48.02 50.79 66.54 57.11 76.36 61.63 70.95 62.21 76.79 69.21 62.22 54.16 61.30 50.79 76.14 52.21 71.68 64.11 66.62 51.47
Qian_VUILabs_task2_1 QianVUILabs2026 55 60.61215647721782 ± 0.0027688037966323107 65.80 59.84 54.03 50.32 55.17 54.47 63.08 50.95 85.24 71.05 78.47 61.63 74.93 62.05 60.97 54.42 59.78 51.84 71.51 49.21 67.04 57.95 66.74 50.47
Qian_VUILabs_task2_2 QianVUILabs2026 78 59.360130042965885 ± 0.0029946372561088157 60.68 53.37 51.39 49.84 61.96 56.95 53.28 51.32 87.30 81.68 60.93 54.05 65.75 55.26 63.61 58.84 58.52 51.89 73.05 55.47 73.34 62.42 69.54 56.84
Qian_VUILabs_task2_3 QianVUILabs2026 56 60.39308038450205 ± 0.002915407352307563 66.68 57.58 52.30 50.11 57.16 57.05 62.48 52.58 82.93 66.00 71.32 64.16 75.54 63.32 65.69 58.74 62.93 50.53 77.87 57.42 73.72 63.89 72.46 57.11
Qian_VUILabs_task2_4 QianVUILabs2026 63 60.14551190647658 ± 0.002983509786380337 62.18 54.26 53.25 50.53 62.37 58.32 54.70 51.21 85.11 79.05 61.07 57.21 67.10 54.58 63.88 60.00 59.69 50.37 72.79 53.68 75.49 64.05 71.52 61.21
Zhang_XJTLU_task2_1 ZhangXJTLU2026 133 55.08676326285066 ± 0.0027795644119063814 53.91 52.84 54.38 49.05 60.96 56.05 55.05 51.68 58.28 54.16 56.56 56.32 65.02 56.21 55.00 56.05 55.98 53.16 68.02 61.42 66.40 55.84 87.70 75.42
Zhang_XJTLU_task2_2 ZhangXJTLU2026 134 55.0843134951928 ± 0.0027805426223342328 53.89 52.84 54.39 49.05 60.96 56.05 55.06 51.68 58.27 54.16 56.56 56.32 65.02 56.21 55.00 56.05 55.98 53.16 68.10 61.26 66.40 55.84 87.70 75.42
Zhang_XJTLU_task2_3 ZhangXJTLU2026 132 55.13731046401534 ± 0.0027395110835357174 54.18 53.11 54.48 49.05 61.21 55.63 54.49 51.26 58.82 54.53 56.56 56.32 65.02 56.21 55.00 56.05 55.96 53.05 67.04 61.05 66.18 55.84 87.70 75.42
Zhang_XJTLU_task2_4 ZhangXJTLU2026 131 55.17563151568782 ± 0.002798601551388066 54.10 53.11 54.53 49.21 61.15 55.84 54.90 51.37 58.65 54.21 56.56 56.32 65.02 56.21 55.00 56.05 55.94 53.05 68.00 57.16 66.40 56.00 87.70 75.42
Chang_Surrey_task2_1 ChangSurrey2026 58 60.32203484103438 ± 0.0026443592346275958 43.66 52.63 61.19 56.11 71.70 55.42 64.19 56.95 80.68 71.89 75.14 52.74 68.80 54.74 65.80 56.21 63.74 52.26 78.58 63.42 82.82 61.37 81.06 59.47
Chang_Surrey_task2_2 ChangSurrey2026 76 59.40285179129806 ± 0.0027963847512381627 49.00 55.21 57.68 51.16 69.75 58.58 60.50 55.16 73.44 65.79 72.78 53.05 66.66 54.21 63.82 56.05 63.28 52.84 75.16 58.05 83.60 61.53 84.70 62.16
Chang_Surrey_task2_3 ChangSurrey2026 112 57.128143744363946 ± 0.002681514157837761 43.12 52.74 59.25 49.79 68.05 60.84 60.26 56.58 68.12 57.68 61.90 53.11 67.22 52.16 57.60 53.37 63.78 55.47 71.96 55.11 82.08 62.68 80.22 67.68
Chang_Surrey_task2_4 ChangSurrey2026 142 53.4290733300099 ± 0.0027290305181672607 41.44 53.89 56.33 55.89 53.97 49.84 49.43 49.05 73.43 65.26 52.94 51.00 63.92 53.74 58.46 53.47 52.70 52.47 59.64 51.74 60.60 54.84 77.16 69.21
Zhang_SATLab_task2_1 ZhangSATLab2026 45 61.102749328726404 ± 0.0027831981617583833 70.19 55.16 61.73 53.32 52.73 52.37 67.25 56.84 74.74 60.26 66.56 57.05 83.19 72.42 63.56 57.16 58.53 50.37 86.23 66.37 74.73 62.68 69.71 53.00
Zhang_SATLab_task2_2 ZhangSATLab2026 71 59.62674106806026 ± 0.0029640774023364097 65.83 60.63 54.37 50.11 51.76 52.11 63.13 53.11 82.76 64.79 79.17 62.84 79.46 66.79 67.22 61.89 64.10 51.53 76.64 51.53 75.38 64.05 79.82 67.53
Zhang_SATLab_task2_3 ZhangSATLab2026 39 61.24923201074941 ± 0.002952301148904951 73.48 60.95 58.59 52.79 49.73 53.00 65.69 56.37 79.71 63.16 73.12 59.58 80.94 71.42 64.78 56.79 61.48 50.95 87.14 67.79 76.55 62.79 74.01 54.11
Zhang_SATLab_task2_4 ZhangSATLab2026 44 61.14285326094995 ± 0.0030181134269168922 69.82 60.58 57.63 52.58 52.69 53.42 63.60 53.42 82.79 64.74 80.89 65.79 81.08 68.74 65.72 60.84 64.39 51.84 82.73 54.21 76.25 63.58 79.43 65.58
Fan_WISTLAB_task2_1 FanWISTLAB2026 35 61.59723652396467 ± 0.0030516512847425276 67.35 62.32 56.87 52.47 61.47 62.26 59.11 52.26 74.47 66.11 68.08 60.58 71.92 54.63 65.84 61.26 62.98 53.32 69.86 51.89 76.40 66.42 77.49 58.05
Fan_WISTLAB_task2_2 FanWISTLAB2026 33 61.65262669097673 ± 0.0030336056906432223 67.14 61.47 57.80 52.68 61.79 60.89 59.71 52.21 75.24 63.95 69.52 58.63 76.64 57.26 66.79 60.95 63.11 53.05 66.83 52.26 80.13 66.21 80.48 59.37
Fan_WISTLAB_task2_3 FanWISTLAB2026 34 61.616388199058115 ± 0.0030055924645254607 67.20 61.79 57.73 52.74 61.81 60.63 59.66 52.11 74.91 64.00 69.34 59.11 76.39 57.11 66.66 60.84 63.42 53.21 66.82 52.47 80.06 66.05 80.26 58.95
Fan_WISTLAB_task2_4 FanWISTLAB2026 17 62.99413985866293 ± 0.003148125130529185 68.27 61.26 59.34 53.63 56.73 57.42 65.15 53.84 83.19 70.58 68.79 60.42 80.68 64.32 65.34 60.32 66.39 51.32 72.03 53.32 81.44 67.11 80.08 56.63
Jiang_AITHU_task2_1 JiangAITHU2026 98 57.69354381091058 ± 0.0026717673355187758 69.02 61.58 48.79 50.16 48.54 52.79 66.18 55.95 69.80 59.00 72.42 61.74 75.30 66.42 61.77 53.79 58.28 50.53 78.27 53.95 69.06 61.32 69.73 53.74
Jiang_AITHU_task2_2 JiangAITHU2026 82 58.899944818447615 ± 0.0028310751297084126 70.34 62.11 49.49 50.21 48.48 51.63 67.18 54.47 78.25 64.53 72.35 61.21 78.70 70.89 63.56 55.63 61.61 51.32 77.94 57.79 72.93 64.11 67.42 51.16
Jiang_AITHU_task2_3 JiangAITHU2026 21 62.3631416226516 ± 0.00304945479733028 72.59 62.16 57.97 52.95 55.67 56.21 63.77 53.47 79.73 68.32 71.80 60.05 80.72 67.74 65.50 59.68 63.48 51.74 81.15 56.58 79.71 67.37 78.93 57.53
Jiang_AITHU_task2_4 JiangAITHU2026 41 61.19087844185517 ± 0.0028886080153183763 73.25 60.32 56.31 51.58 51.10 52.42 64.99 56.16 80.63 67.84 71.21 59.68 80.25 72.05 64.37 55.42 61.73 50.42 87.56 70.89 75.08 62.16 74.81 54.63
Zhang_THUEE_task2_1 ZhangTHUEE2026 30 61.87003460995337 ± 0.0030168009209201752 67.62 59.47 55.59 52.05 58.42 55.47 61.66 54.26 86.23 68.79 66.94 55.21 79.13 66.68 63.07 60.11 65.01 51.58 83.11 66.53 76.03 64.63 71.80 55.37
Zhang_THUEE_task2_2 ZhangTHUEE2026 51 60.82146453318818 ± 0.0030246061291548656 70.89 60.89 56.30 51.89 52.06 52.53 64.07 53.89 82.33 63.79 80.44 64.79 81.56 69.42 65.64 61.37 64.47 51.89 83.34 54.53 75.59 63.63 78.98 64.79
Zhang_THUEE_task2_3 ZhangTHUEE2026 24 62.12111242397091 ± 0.00309980035586224 72.40 62.74 57.12 53.21 54.06 55.58 63.87 53.00 83.18 67.21 79.23 63.26 81.37 67.16 66.50 61.16 64.93 51.95 81.98 54.00 78.31 65.68 80.59 64.58
Zhang_THUEE_task2_4 ZhangTHUEE2026 11 63.71007071087524 ± 0.003109602589321203 69.98 62.37 59.67 54.89 59.95 60.05 63.63 52.79 83.14 68.26 69.74 58.11 79.37 62.79 65.89 61.21 65.37 52.00 76.63 53.58 81.13 70.84 79.34 58.05
Huang_CQUPT_task2_1 HuangCQUPT2026 111 57.22358595466307 ± 0.002965303513875292 55.16 54.89 47.38 50.26 65.19 65.95 55.31 50.84 69.18 66.84 54.79 51.37 60.21 52.89 63.53 61.26 67.43 56.84 68.26 50.11 80.35 62.11 85.87 79.53
Huang_CQUPT_task2_2 HuangCQUPT2026 124 56.31601728701303 ± 0.0028354657152074767 50.93 56.26 46.70 49.89 64.55 64.89 56.00 51.26 67.37 66.05 53.62 50.00 61.87 50.26 57.71 56.21 67.08 52.95 68.31 48.47 79.13 61.11 78.28 64.05
Huang_CQUPT_task2_3 HuangCQUPT2026 122 56.80273782327397 ± 0.002854832632939019 52.39 55.63 47.23 50.21 64.96 66.58 55.57 50.95 68.21 67.05 54.19 51.00 61.14 52.05 61.92 60.63 67.59 55.11 68.47 49.32 80.36 63.68 82.83 73.58
Huang_CQUPT_task2_4 HuangCQUPT2026 109 57.345977001988956 ± 0.003004740948360317 58.20 56.53 45.14 49.95 65.05 62.63 54.71 50.68 72.46 68.53 57.43 51.26 61.56 54.32 59.37 58.68 69.05 55.58 66.48 50.05 78.24 57.95 83.99 74.84
Xie_SHU_task2_1 XieSHU2026 75 59.4198374590017 ± 0.0028492593511631223 49.50 50.79 54.97 52.21 71.27 65.47 63.43 61.26 70.79 57.58 50.86 53.26 64.15 52.05 62.66 54.47 70.86 60.05 68.89 57.21 75.16 60.37 82.70 72.21
Xie_SHU_task2_2 XieSHU2026 65 60.03523389769877 ± 0.002807957195771089 51.20 51.26 56.77 51.89 69.51 61.68 64.30 61.84 73.18 58.16 53.00 53.53 65.46 52.21 62.49 55.00 72.02 59.79 69.74 57.05 74.60 60.11 83.13 70.89
Xie_SHU_task2_3 XieSHU2026 84 58.71898619396716 ± 0.0028101255197292936 49.72 48.95 56.41 54.05 65.89 62.00 63.01 61.00 69.16 57.79 57.01 50.47 63.16 51.58 61.04 54.63 67.74 55.16 69.46 55.05 74.19 57.47 87.21 79.68
Xie_SHU_task2_4 XieSHU2026 93 57.85067191322667 ± 0.0027825271036808147 50.05 48.53 55.98 53.47 62.90 59.32 62.36 61.16 67.81 56.42 57.46 52.21 63.69 52.74 61.16 54.21 68.01 52.37 67.35 55.84 73.51 57.89 87.86 80.21
Moon_Independent_task2_1 MoonIndependent2026 127 55.60880820594806 ± 0.002826442080110754 54.68 55.00 52.37 54.26 62.32 60.00 56.04 50.79 58.54 49.84 64.66 50.58 70.72 64.00 63.12 60.42 62.20 53.42 79.26 57.37 70.14 54.58 90.16 83.21
Xia_NEU_task2_1 XiaNEU2026 32 61.72131772891849 ± 0.002899363535449049 63.93 53.84 50.03 50.68 68.34 69.16 58.25 51.47 86.85 77.21 63.90 53.43 74.26 57.45 56.90 49.78 75.68 55.60 81.18 50.22 73.24 57.29 92.46 84.75
Xia_NEU_task2_2 XiaNEU2026 69 59.731876891962365 ± 0.002932783413517079 55.97 59.11 55.34 51.26 70.72 70.00 54.29 50.37 73.68 60.89 54.00 53.73 68.90 53.91 55.88 55.53 62.58 51.15 71.90 51.70 78.88 57.52 91.22 68.32
Xia_NEU_task2_3 XiaNEU2026 67 59.83019547452064 ± 0.002926768054379842 55.38 59.00 52.41 51.53 71.27 68.47 54.19 50.47 77.95 66.53 54.74 55.46 71.04 55.43 55.38 55.83 63.26 50.90 72.06 50.28 77.98 53.29 90.60 67.76
Xia_NEU_task2_4 XiaNEU2026 46 61.06855783561832 ± 0.0029221942639870695 60.23 59.47 57.42 52.95 70.72 67.26 53.26 50.16 77.16 65.21 52.80 54.63 70.32 53.62 56.80 54.34 63.34 49.76 71.02 48.49 77.68 55.19 93.66 74.49
XingWu_MCPX_task2_1 XingWuMCPX2026 50 60.953892385175514 ± 0.0027430558585857705 62.10 56.63 53.56 49.89 65.74 61.21 59.76 52.32 80.50 68.26 64.96 53.42 66.82 59.89 58.18 55.42 55.16 48.05 71.34 49.63 79.08 57.74 63.86 51.89
XingWu_MCPX_task2_2 XingWuMCPX2026 36 61.5341202167328 ± 0.0027605246868952237 62.55 58.26 54.51 50.21 66.23 62.16 60.05 52.16 80.63 69.21 67.18 54.21 68.52 59.68 58.68 55.05 57.04 48.89 70.46 50.58 79.34 57.32 66.30 52.16
XingWu_MCPX_task2_3 XingWuMCPX2026 74 59.478432884095554 ± 0.002730362813381321 60.20 54.47 53.09 51.95 60.72 59.37 58.27 48.84 79.99 70.89 50.94 50.32 66.80 58.37 57.72 54.32 52.94 48.21 68.44 50.74 65.80 52.84 65.18 50.84
XingWu_MCPX_task2_4 XingWuMCPX2026 29 61.87465528410395 ± 0.002766292907182805 63.29 57.84 57.38 51.37 65.61 61.47 59.63 51.84 79.51 69.05 67.08 54.11 68.08 59.11 58.74 54.58 56.18 48.68 70.26 52.53 77.86 56.53 67.12 52.11
Zhou_SUMERUZOO_task2_1 ZhouSUMERUZOO2026 175 48.36326455107135 ± 0.002487223795828334 34.46 52.16 48.44 50.32 52.76 51.11 53.16 51.68 52.46 52.05 50.97 51.58 60.23 50.47 56.14 55.42 47.13 50.58 54.27 53.47 62.15 53.53 70.44 54.26
Zhou_SUMERUZOO_task2_3 ZhouSUMERUZOO2026 174 48.43662010700282 ± 0.002323678592788652 40.43 51.00 48.91 49.11 49.73 47.58 50.01 50.11 51.60 51.53 68.36 55.53 52.99 49.32 63.63 53.63 61.00 55.16 58.00 56.53 56.01 52.79 69.01 61.05
Kwon_KIST_task2_1 KwonKIST2026 110 57.33672495384062 ± 0.0028127046901336617 46.02 50.11 52.08 52.89 57.60 62.00 61.68 53.05 80.05 72.58 69.70 61.42 72.10 61.11 68.40 61.32 67.18 56.74 75.10 52.47 80.68 63.95 91.96 78.42
Kwon_KIST_task2_2 KwonKIST2026 114 57.088119822300875 ± 0.002866676336007036 45.53 49.89 52.50 52.79 57.15 62.53 61.38 53.58 78.79 70.79 69.12 60.84 70.98 60.89 67.20 60.21 65.48 55.58 74.06 53.16 79.68 62.11 91.06 77.47
Kwon_KIST_task2_3 KwonKIST2026 120 56.81009796912044 ± 0.0028425070155682825 44.73 49.84 52.53 52.84 57.02 62.37 61.24 53.58 78.07 70.37 68.66 60.47 71.08 60.32 67.16 59.95 65.50 55.53 73.98 53.58 79.02 61.47 90.60 76.63
Kwon_KIST_task2_4 KwonKIST2026 113 57.11119213398058 ± 0.0028474676672952653 45.70 49.84 52.62 52.84 57.20 62.58 61.18 53.63 78.54 70.68 68.70 60.84 71.06 60.79 67.16 60.21 65.50 55.58 74.08 53.16 79.30 61.95 90.80 76.89
Fujimura_MERL_task2_1 FujimuraMERL2026 2 66.00368477798287 ± 0.0031542001840882444 68.37 63.21 62.45 59.26 67.84 65.58 60.58 54.05 85.30 73.53 62.03 53.16 72.02 57.16 59.39 58.74 77.77 56.74 74.54 50.84 77.94 59.68 91.20 86.84
Fujimura_MERL_task2_2 FujimuraMERL2026 15 63.21724463593549 ± 0.0028967420338867874 67.39 59.68 53.93 51.00 65.36 63.74 64.68 53.63 80.64 74.11 57.45 52.47 71.67 53.21 64.43 59.74 69.95 60.47 69.60 51.00 66.72 56.26 94.35 84.26
Fujimura_MERL_task2_3 FujimuraMERL2026 1 70.24115096031407 ± 0.0030780430693608625 71.79 63.53 68.36 61.74 71.45 68.79 69.52 57.37 88.64 75.74 57.58 51.21 77.45 58.00 67.15 62.05 75.33 69.42 70.59 51.05 65.32 52.00 98.20 91.89
Fujimura_MERL_task2_4 FujimuraMERL2026 6 64.99707422230968 ± 0.002842021913428696 68.32 61.89 52.53 50.26 68.15 60.47 68.10 60.05 87.66 79.00 63.91 53.11 74.23 60.26 64.95 57.79 71.01 58.11 69.77 49.58 65.65 56.68 92.58 85.84
Noh_CBNU_task2_1 NohCBNU2026 118 56.88833422457569 ± 0.002686961640274201 58.01 54.47 52.33 53.53 61.84 57.21 55.82 54.32 65.75 51.37 53.99 54.16 72.77 62.26 65.95 59.37 57.73 53.08 67.81 54.05 65.97 52.29 89.32 76.05
Jeong_KETI_task2_1 JeongKETI2026 130 55.25280412349333 ± 0.0027578476609583817 51.41 55.05 58.24 51.00 49.51 50.58 58.25 56.05 63.74 59.74 42.86 50.58 62.00 56.53 58.53 53.37 67.43 56.42 69.13 56.84 61.34 51.63 53.48 52.79
Zarrouky_IR_task2_1 ZarroukyIR2026 140 53.833412450926346 ± 0.0025414134115744453 59.02 51.68 48.80 50.13 54.33 51.34 48.17 48.37 70.14 57.26 91.98 73.05 52.73 50.53 60.54 57.08 60.06 50.97 57.28 51.84 65.42 52.79 76.06 54.00
Lei_CRRC_task2_1 LeiCRRC2026 136 54.9033413783018 ± 0.0026858581106864094 59.83 58.58 43.36 53.58 51.19 50.89 53.96 50.11 75.38 63.89 54.70 52.00 64.00 55.16 65.27 59.84 57.79 50.68 67.64 54.42 65.60 57.11 55.97 49.26
Lei_CRRC_task2_2 LeiCRRC2026 128 55.44466150877836 ± 0.0025805325640264848 58.51 59.47 43.95 52.74 53.84 52.00 54.46 50.32 75.46 64.11 54.27 52.89 63.90 55.16 64.02 59.63 57.65 50.74 66.99 53.58 67.05 57.32 60.38 49.68
Lei_CRRC_task2_3 LeiCRRC2026 149 52.78863428349989 ± 0.002564685552316373 55.53 57.84 37.91 49.21 49.99 49.95 54.00 50.47 77.97 65.63 55.67 51.42 62.50 53.26 63.44 60.89 57.39 50.95 64.05 53.53 67.51 55.58 58.63 49.16
Lei_CRRC_task2_4 LeiCRRC2026 108 57.38198513172906 ± 0.0028700109628889314 44.14 50.89 56.21 51.58 66.27 62.42 59.52 54.68 72.45 63.37 59.55 52.89 71.84 62.47 55.73 60.16 68.22 56.95 67.64 50.05 79.64 62.47 78.85 58.37
Tsz_HFU_task2_1 TszHFU2026 126 55.94984660573674 ± 0.0028534107013166777 59.93 58.63 47.27 52.74 50.86 50.74 55.69 49.84 77.30 64.47 53.01 52.89 64.02 53.16 65.01 60.47 58.09 51.16 70.46 53.16 65.30 57.74 57.81 50.68
Tsz_HFU_task2_2 TszHFU2026 156 51.253376236546174 ± 0.0025870928096575134 39.54 52.13 55.41 50.13 56.33 52.05 51.66 51.66 56.19 54.16 52.62 53.11 72.66 63.32 67.52 59.37 64.88 56.42 71.44 48.84 67.68 48.74 70.88 61.53
Tsz_HFU_task2_3 TszHFU2026 103 57.51642313488529 ± 0.0027109071470683395 61.22 50.79 60.00 53.26 60.70 55.58 57.79 53.84 61.52 50.79 44.62 50.11 68.47 55.00 53.67 52.42 56.34 49.16 67.67 52.05 64.57 49.42 72.72 55.95
Tsz_HFU_task2_4 TszHFU2026 170 49.81796061059645 ± 0.0022782640230914987 54.20 51.95 42.73 49.53 41.77 47.89 58.10 55.11 54.25 50.63 57.88 59.89 61.89 53.63 60.48 53.05 51.31 48.63 54.80 49.74 56.34 51.37 47.41 48.95
Kim_LUDO_task2_1 KimLUDO2026 70 59.69660936511764 ± 0.0029047720393691823 57.60 63.47 51.53 52.79 64.71 62.89 54.60 51.26 76.82 71.11 48.97 52.89 62.46 52.95 64.64 60.16 75.13 55.53 69.98 49.47 81.00 61.79 82.51 71.84
Kim_LUDO_task2_2 KimLUDO2026 64 60.10940314826085 ± 0.0029819859022725327 57.38 62.74 51.74 52.63 64.03 63.68 55.88 52.00 78.48 73.26 51.72 52.89 64.83 52.95 63.79 60.53 73.09 55.05 71.50 50.00 82.43 59.47 76.44 56.74
Kim_LUDO_task2_3 KimLUDO2026 53 60.72626822655016 ± 0.002906834203180997 58.06 63.47 55.78 54.58 64.36 63.42 53.28 51.68 79.35 73.00 49.04 52.95 62.73 53.11 61.80 59.95 71.00 55.37 70.48 49.74 81.21 60.26 81.09 69.00
Kim_LUDO_task2_4 KimLUDO2026 102 57.55294726230864 ± 0.002757097225349844 52.70 52.37 51.91 49.21 66.65 63.37 55.58 51.74 71.34 64.11 55.24 53.21 62.00 53.74 58.55 59.32 61.67 51.95 71.30 58.11 71.80 61.63 64.26 52.58
Balozi_RISE_task2_1 BaloziRISE2026 155 51.266508314982 ± 0.002535299411533062 36.93 50.47 50.78 48.53 59.80 59.42 52.60 49.68 61.92 55.84 51.01 48.95 65.61 58.11 53.33 52.11 59.66 52.63 60.95 52.26 71.89 63.03 76.47 56.32
Balozi_RISE_task2_2 BaloziRISE2026 154 51.297542290502165 ± 0.0025676312077536315 34.96 49.74 52.06 48.95 60.60 60.05 53.25 50.68 64.55 54.58 49.39 49.89 63.11 56.95 52.97 52.11 57.88 53.89 61.62 51.95 71.16 58.84 75.97 55.32
Balozi_RISE_task2_3 BaloziRISE2026 166 50.6952981490369 ± 0.002559605378173619 33.96 49.74 51.09 48.89 60.07 60.37 53.46 49.21 63.49 54.95 51.16 48.74 65.41 60.58 53.15 51.79 58.73 53.95 61.59 50.63 72.03 61.32 75.92 54.95
Balozi_RISE_task2_4 BaloziRISE2026 157 51.247843521934286 ± 0.002586957976870847 36.70 51.11 50.44 48.53 60.59 59.42 52.09 49.68 62.47 55.58 50.62 49.32 65.80 60.26 53.04 51.89 60.58 52.84 61.01 52.16 70.79 62.37 75.88 55.68
Mei_FDID_task2_1 MeiFDID2026 129 55.39000249761615 ± 0.0026175082253723754 55.07 50.68 49.14 50.53 66.13 58.89 52.74 49.53 63.83 56.84 81.75 61.40 60.10 54.00 62.40 60.00 59.60 50.70 75.00 58.20 75.25 55.10 85.90 61.70
Mei_FDID_task2_2 MeiFDID2026 138 54.51541523323363 ± 0.0026057255894381 53.24 50.74 47.95 50.58 59.02 56.05 57.18 49.53 65.21 53.11 80.70 59.60 60.85 54.20 64.00 59.80 61.75 51.00 74.90 58.30 74.85 54.90 85.90 62.20
Mei_FDID_task2_3 MeiFDID2026 150 52.52626407141045 ± 0.0025020612183845083 48.26 50.79 50.25 50.79 55.80 50.74 52.42 48.32 64.86 50.74 70.50 55.50 61.00 54.30 62.90 59.90 54.80 49.10 74.60 58.20 74.65 54.80 84.65 61.10
Wang_WST_task2_1 WangWST2026 148 52.846839306143245 ± 0.0026793149709115026 51.07 53.16 45.39 51.68 51.19 50.95 48.12 49.63 74.87 65.21 62.84 52.53 68.32 55.84 64.76 61.11 57.22 50.42 74.28 57.95 72.42 56.26 69.10 53.58
Wang_WST_task2_2 WangWST2026 173 48.92744575600344 ± 0.002513331730611134 32.11 48.74 54.75 49.47 54.75 52.11 48.15 49.26 63.89 55.05 63.42 62.37 57.18 56.77 58.08 58.29 52.10 52.38 66.48 65.30 64.44 57.81 62.30 62.41
Wang_WST_task2_3 WangWST2026 145 53.13694610777225 ± 0.00274928682418755 49.31 53.05 46.28 50.84 53.56 53.16 48.39 49.68 74.71 64.05 63.20 52.42 66.40 56.32 63.74 61.63 56.06 50.11 73.26 57.58 70.96 54.42 67.72 53.74
Jiang_KY_task2_1 JiangKY2026 167 50.432955685562284 ± 0.0023957097828267333 48.95 53.68 46.00 49.58 47.54 49.37 48.19 47.95 64.56 53.68 70.52 62.63 65.70 52.47 61.02 52.05 49.66 52.42 39.50 47.37 45.48 49.42 50.80 48.68
Jiang_KY_task2_2 JiangKY2026 160 51.165463062186866 ± 0.0023842000163354577 51.49 54.84 50.34 49.58 50.22 49.11 47.33 47.84 60.90 50.00 73.36 62.53 65.78 52.63 62.58 53.00 42.28 52.16 39.50 47.37 45.48 49.42 54.54 49.11
Jiang_KY_task2_3 JiangKY2026 141 53.558263051309616 ± 0.002623267856524193 53.13 52.26 51.86 48.89 48.87 51.74 56.61 50.21 66.78 53.11 70.56 54.47 73.12 60.79 56.52 49.32 50.20 49.84 41.66 47.68 45.88 50.42 50.46 51.11
Jiang_KY_task2_4 JiangKY2026 144 53.17379842095369 ± 0.002546961717770218 50.56 51.79 53.91 49.42 47.23 50.84 57.01 50.26 67.31 52.11 70.54 53.79 72.64 60.42 56.28 51.53 51.38 51.42 41.66 47.68 45.88 50.42 51.58 48.47
Glitza_IKA_task2_1 GlitzaIKA2026 153 51.404261059460744 ± 0.0024366142866778908 41.93 50.79 53.26 50.76 55.16 50.37 50.96 48.26 62.55 52.32 57.05 52.26 62.07 55.68 56.64 53.89 49.66 48.42 72.97 51.32 66.17 52.42 76.25 68.21
Glitza_IKA_task2_2 GlitzaIKA2026 147 52.85572485680369 ± 0.002592134338288459 44.01 52.53 54.23 50.74 56.46 55.95 50.01 49.42 64.56 54.89 56.26 54.05 58.73 53.53 58.62 51.79 53.20 48.74 69.01 59.11 57.69 55.11 78.23 69.42
Glitza_IKA_task2_3 GlitzaIKA2026 139 54.4613765901715 ± 0.0025487340237951833 54.50 50.63 47.73 51.53 56.92 53.71 56.39 50.26 67.43 52.95 57.42 51.68 58.36 49.37 60.66 59.16 58.20 48.79 56.10 53.47 59.53 51.00 44.24 49.34
Glitza_IKA_task2_4 GlitzaIKA2026 164 50.772616270618464 ± 0.002455376130479949 48.88 51.63 47.51 49.84 53.26 48.47 49.65 50.68 58.61 48.37 52.45 50.79 57.60 48.11 54.36 53.21 53.40 48.37 60.99 54.79 48.91 50.79 31.83 48.53
Kajita_IND_task2_1 KajitaIND2026 159 51.16968943337452 ± 0.0024973846824596496 55.44 53.63 48.49 49.84 52.97 51.11 50.01 49.84 50.18 50.32 49.24 51.37 56.73 50.84 52.78 50.95 48.03 50.11 51.87 50.05 45.59 50.05 50.91 49.37
Kajita_IND_task2_2 KajitaIND2026 161 51.12981391663411 ± 0.002477359370609657 56.39 53.68 47.26 49.76 52.69 51.32 50.18 50.21 50.28 50.32 48.62 50.63 55.25 51.63 53.51 50.05 48.25 49.84 53.43 50.47 44.93 48.47 49.72 49.68
Kajita_IND_task2_3 KajitaIND2026 151 51.77972417170499 ± 0.00243164973546122 60.64 53.58 48.93 49.76 53.20 51.53 52.35 50.11 47.61 50.26 56.38 50.58 53.32 51.26 53.13 50.21 47.08 49.89 56.05 50.58 45.83 48.37 48.01 49.95
Kajita_IND_task2_4 KajitaIND2026 158 51.18773685083791 ± 0.002469901606998131 55.71 53.79 48.08 49.95 52.98 51.05 50.21 49.89 50.19 50.32 48.30 50.47 56.67 51.05 53.08 50.53 48.21 49.95 52.75 50.47 44.41 48.26 50.84 50.11
Zhang_JAIST_task2_1 ZhangJAIST2026 47 61.02577712684993 ± 0.0030562402643252644 57.52 54.42 60.15 51.00 67.28 64.37 57.56 53.47 75.21 69.00 64.61 54.74 70.55 62.58 64.56 60.42 70.90 58.58 69.07 53.63 79.00 58.79 79.77 64.53
Wang_UniS_task2_1 WangUniS2026 72 59.593310968513094 ± 0.0027704226868877947 59.58 56.68 53.05 49.05 58.77 57.63 55.90 52.16 86.01 78.00 63.36 52.42 70.32 53.21 64.00 60.00 63.56 49.16 78.04 60.00 76.10 61.84 81.16 68.26
Wang_UniS_task2_2 WangUniS2026 143 53.384478032984305 ± 0.002475972003618646 36.14 48.53 42.32 49.26 66.96 65.53 57.91 51.95 86.23 65.11 56.64 47.84 71.00 59.26 63.48 60.11 66.36 54.37 75.46 49.74 77.16 59.79 81.22 75.42
Wang_UniS_task2_3 WangUniS2026 87 58.40472656042925 ± 0.0029863899547199852 50.95 54.42 49.21 49.11 62.59 62.05 57.05 50.63 89.55 77.08 62.69 52.18 72.21 54.47 65.48 61.42 67.68 52.95 78.08 55.21 79.32 62.66 84.97 78.89
Wang_UniS_task2_4 WangUniS2026 92 57.96828640077412 ± 0.0029184901554340837 49.35 54.21 48.48 48.97 62.93 62.24 57.10 50.61 89.66 75.95 62.64 51.84 72.31 55.00 65.45 61.63 67.63 53.03 78.10 54.92 79.22 62.24 84.94 78.89
Yang_XJU_task2_1 YangXJU2026 99 57.68682737042624 ± 0.00286598202217133 50.10 55.47 46.03 48.42 65.75 61.11 63.14 52.84 79.42 64.89 61.80 49.11 70.84 58.74 62.81 58.11 67.12 52.00 78.24 61.11 75.09 61.21 71.21 64.00
Yang_XJU_task2_2 YangXJU2026 85 58.62426327658036 ± 0.002840695514354596 53.62 50.79 54.52 49.21 61.65 61.58 60.34 51.11 83.00 61.05 55.12 51.05 71.40 61.00 61.46 60.53 55.54 49.74 71.74 55.68 76.86 65.16 79.05 69.21
Yang_XJU_task2_3 YangXJU2026 105 57.47157281227988 ± 0.002829302588588755 48.79 53.95 44.79 48.68 66.04 61.74 65.44 52.21 79.16 67.53 62.48 49.26 70.89 58.16 62.76 57.00 65.68 51.79 78.95 59.58 72.50 56.79 70.16 64.53
Yang_XJU_task2_4 YangXJU2026 62 60.15236729958677 ± 0.002637242776897266 55.01 52.21 51.72 48.21 63.68 62.95 63.78 52.68 86.28 73.37 59.47 51.47 76.32 59.26 62.53 60.32 62.90 50.79 77.64 62.21 79.94 66.58 79.65 73.79
SNU_task2_1 SNUtask22026 169 50.02291439923236 ± 0.0026525044143756536 40.88 51.79 48.09 48.95 44.94 49.79 55.97 53.42 60.84 56.53 61.56 54.37 65.14 55.58 61.66 51.47 61.18 53.47 74.20 59.16 77.92 50.53 83.86 63.32
Huang_WHU_task2_1 HuangWHU2026 4 65.45291026592781 ± 0.003042006671244826 69.29 58.89 64.22 55.16 65.60 65.42 63.84 55.63 80.24 71.16 74.08 53.89 78.21 54.58 74.85 64.95 61.99 52.79 73.48 52.21 80.07 61.11 79.35 52.84
Huang_WHU_task2_2 HuangWHU2026 7 64.96454007817948 ± 0.00293218905783368 69.03 57.74 64.58 54.84 65.24 65.58 63.12 55.16 79.75 68.47 70.58 52.58 77.36 55.00 70.06 60.63 61.47 52.74 73.64 51.68 80.71 60.95 79.87 52.26
Huang_WHU_task2_3 HuangWHU2026 31 61.85076889858421 ± 0.0029609118335136497 69.42 59.18 44.72 51.42 65.33 64.95 62.74 53.50 86.47 77.32 60.77 52.37 69.42 53.47 64.12 62.10 68.91 53.95 75.46 52.79 80.64 56.26 76.93 52.68
Huang_WHU_task2_4 HuangWHU2026 83 58.77264143329869 ± 0.0027920967460074568 53.27 55.37 62.04 52.00 64.37 61.68 60.10 50.16 64.76 59.63 53.10 51.05 69.06 54.26 61.44 61.11 64.80 55.16 68.06 51.79 66.23 51.84 79.36 67.89
Morita_KM_task2_1 MoritaKM2026 117 56.95558037093822 ± 0.0027328498622641033 48.96 53.11 53.27 53.53 59.08 52.79 59.92 52.79 77.09 61.00 68.15 58.68 77.07 56.58 56.78 53.79 66.96 54.16 74.99 55.05 70.34 62.84 80.61 62.53
Morita_KM_task2_2 MoritaKM2026 96 57.75426708264063 ± 0.002834402508879935 50.10 52.00 52.22 52.74 59.10 53.89 59.39 54.16 83.53 67.42 70.09 57.32 78.13 65.26 53.45 52.74 64.68 52.89 73.23 52.11 70.44 61.84 81.05 66.00
Morita_KM_task2_3 MoritaKM2026 97 57.72174786333586 ± 0.00286455045014039 49.60 51.86 55.08 52.58 62.02 55.13 58.22 53.56 76.11 65.61 73.79 51.89 65.92 55.79 60.91 59.89 62.68 53.45 70.60 56.50 65.50 56.95 76.78 54.50
Morita_KM_task2_4 MoritaKM2026 90 58.11117958443154 ± 0.0027940480488639475 54.92 51.74 57.97 52.58 61.25 51.84 57.26 50.58 75.11 64.16 78.03 55.00 65.11 56.21 62.58 60.68 61.91 52.58 68.84 58.53 62.37 56.63 74.01 53.37
Wu_CUMT_task2_1 WuCUMT2026 13 63.2915482865626 ± 0.0031625614652688117 59.52 56.84 56.35 52.37 70.90 60.05 61.96 53.37 89.71 76.68 79.57 61.32 74.89 64.47 58.48 53.68 58.55 49.26 75.89 57.84 69.38 59.58 67.78 58.00
Wu_CUMT_task2_2 WuCUMT2026 9 64.38270902115202 ± 0.003234469276470156 62.47 58.16 61.50 50.89 69.37 60.21 63.40 53.95 89.53 72.79 76.73 58.95 75.90 58.37 60.93 54.21 58.15 49.37 76.27 58.42 75.21 58.05 70.55 61.42
Wu_CUMT_task2_3 WuCUMT2026 5 65.18092078146877 ± 0.003133805101200734 63.91 58.58 64.90 55.42 70.81 61.37 60.03 53.32 90.40 70.58 82.19 62.79 78.91 74.00 59.56 52.63 56.81 49.21 79.03 58.21 67.72 58.47 70.79 54.53
Wu_CUMT_task2_4 WuCUMT2026 3 65.46214466728418 ± 0.0032688205174339403 64.17 57.05 63.74 56.16 69.51 59.05 63.16 55.89 91.32 70.95 78.88 60.53 79.55 65.68 61.89 53.79 59.49 49.63 79.75 60.68 76.71 62.05 74.89 58.16
Yang_None_task2_1 YangNone2026 10 64.29601690030654 ± 0.0031399217014158464 65.01 56.26 58.39 50.95 68.50 58.84 62.76 55.68 90.32 78.63 61.06 60.58 79.18 64.21 58.57 51.74 54.65 49.21 79.17 61.53 73.33 63.11 78.52 67.95
Yang_None_task2_2 YangNone2026 14 63.24325016689553 ± 0.0031265284694510795 61.24 55.42 54.72 51.53 68.37 60.16 62.60 55.89 90.74 78.32 58.75 58.95 77.31 66.26 55.85 52.00 55.02 48.53 78.72 62.26 70.50 62.05 74.62 63.00
Yang_None_task2_3 YangNone2026 16 63.044509254666266 ± 0.002943420835254385 69.17 64.53 50.71 51.11 67.26 59.21 62.82 55.47 87.27 66.21 61.51 56.26 75.18 57.95 57.64 52.21 56.93 49.53 81.87 61.47 76.41 58.74 70.10 56.63
Yang_None_task2_4 YangNone2026 18 62.61820600308388 ± 0.003131768456452344 67.90 63.53 49.24 49.53 67.82 59.68 60.25 53.05 89.96 76.79 61.84 57.42 75.00 55.26 55.26 50.89 56.58 48.89 81.09 64.11 75.45 60.11 66.77 59.84
Zheng_HFUUAI_task2_1 ZhengHFUUAI2026 48 61.023513979163354 ± 0.002849768435747041 68.77 58.95 57.09 53.00 57.60 51.37 59.25 52.95 82.03 68.26 67.48 50.47 72.49 55.05 60.58 59.53 55.16 48.95 66.87 55.37 74.88 62.26 84.78 62.32
Zheng_HFUUAI_task2_2 ZhengHFUUAI2026 40 61.229272315799086 ± 0.002884454716609197 64.71 56.47 58.91 53.84 59.20 53.68 60.39 54.05 80.86 66.32 64.15 51.74 72.69 58.05 58.06 58.79 56.60 49.05 67.17 55.26 74.57 63.37 85.57 65.00
Zheng_HFUUAI_task2_3 ZhengHFUUAI2026 25 62.113376977891996 ± 0.0028591933097166624 68.39 58.84 62.36 51.63 61.53 56.05 59.83 52.68 76.98 66.32 67.51 50.84 72.31 54.79 60.58 60.11 55.14 48.84 65.76 55.00 74.88 62.47 84.76 61.95
Zheng_HFUUAI_task2_4 ZhengHFUUAI2026 123 56.47889079871678 ± 0.002821361575879175 64.66 59.37 46.53 54.32 50.32 49.32 55.66 49.84 78.92 66.42 64.12 51.95 72.45 57.79 58.07 59.00 56.68 48.84 66.35 54.89 74.50 62.21 85.60 64.74
Zhou_HFUUDS_task2_1 ZhouHFUUDS2026 28 61.94635540431729 ± 0.0028420202354601303 68.87 58.21 61.00 51.74 61.56 56.47 59.24 52.89 76.96 66.84 62.94 53.68 72.92 57.79 61.07 60.84 57.52 49.47 66.05 53.00 77.07 62.47 87.37 69.11
Zhou_HFUUDS_task2_2 ZhouHFUUDS2026 22 62.329276538578405 ± 0.002881993367044493 67.88 57.26 59.85 52.11 62.01 56.89 60.98 53.79 79.37 68.00 63.06 53.95 70.64 57.53 60.53 60.89 57.43 49.63 66.39 53.00 77.30 62.53 87.26 69.16
Zhou_HFUUDS_task2_3 ZhouHFUUDS2026 26 62.09915030625375 ± 0.0028509064842366584 68.50 58.84 61.92 51.79 61.40 56.42 59.87 52.79 76.66 66.58 66.61 50.89 72.34 54.68 60.30 60.32 55.64 48.95 65.94 54.74 75.33 62.74 84.92 62.58
Zhou_HFUUDS_task2_4 ZhouHFUUDS2026 20 62.4013149311637 ± 0.0028683405358712537 67.63 57.37 60.31 52.05 61.95 56.74 60.92 53.74 79.98 68.26 64.94 51.79 71.97 55.84 58.94 60.68 56.70 49.26 65.85 54.26 76.21 62.00 86.54 67.11
Zhu_FDA_task2_1 ZhuFDA2026 95 57.78164166933277 ± 0.0028994612385944783 51.41 50.37 50.90 51.68 61.32 59.16 59.30 53.89 75.26 72.37 64.90 51.95 64.70 58.05 65.70 58.00 68.51 52.47 64.78 58.11 61.30 54.11 83.37 74.37
Zhu_FDA_task2_2 ZhuFDA2026 115 57.08274915208862 ± 0.002884203012621284 49.74 50.11 50.07 51.26 61.02 59.11 59.18 53.89 74.29 70.47 65.36 52.26 64.49 57.74 65.37 57.74 67.50 52.32 64.41 59.42 58.57 54.95 82.72 72.79
Zhu_FDA_task2_3 ZhuFDA2026 100 57.660595219017964 ± 0.0029031460352465883 53.36 51.21 51.78 51.84 61.68 59.11 56.70 52.37 75.65 66.37 63.91 52.79 61.38 53.11 62.56 56.21 65.94 52.58 66.50 60.89 60.53 51.89 82.54 66.58
Zhu_FDA_task2_4 ZhuFDA2026 94 57.81527446524864 ± 0.0028700545499872093 50.49 50.32 51.30 51.74 62.09 59.89 59.72 54.16 74.40 71.95 64.33 52.37 65.33 57.63 65.80 57.47 67.30 52.42 63.73 57.68 60.34 54.05 81.01 67.16
Huang_QWS_task2_1 HuangQWS2026 27 62.0432723795996 ± 0.002884666717481283 59.84 53.74 56.96 49.68 65.46 64.05 61.25 55.26 85.98 69.84 64.38 58.00 72.19 63.32 66.10 59.05 64.69 53.58 63.22 55.53 69.28 55.61 92.86 80.63
Wang_Liu_SuzhouDongyuan_task2_1 WangLiu2026 146 53.07523562924925 ± 0.002651215737798937 50.61 51.84 44.91 49.42 55.24 55.32 55.86 50.37 65.33 54.11 64.36 52.37 67.52 53.21 57.60 55.47 52.68 48.79 69.82 55.26 77.26 63.68 79.08 59.37
Wang_Liu_SuzhouDongyuan_task2_2 WangLiu2026 163 50.89827291845622 ± 0.0025143754654380866 47.49 52.63 45.37 49.00 50.24 50.79 55.19 50.63 57.76 52.68 68.28 53.16 67.70 52.95 56.38 53.37 52.18 48.16 68.46 54.68 77.48 63.05 78.68 60.26
Wang_Liu_SuzhouDongyuan_task2_3 WangLiu2026 168 50.20847018027899 ± 0.002508676462631817 47.46 52.95 45.73 49.11 50.03 50.79 54.83 50.58 52.51 50.95 63.34 52.16 67.06 53.11 57.78 55.37 52.66 48.79 69.42 55.95 77.12 63.58 78.98 59.42
Wang_Liu_SuzhouDongyuan_task2_4 WangLiu2026 152 51.58814514961377 ± 0.0024870486076539037 49.31 51.26 44.87 49.58 55.40 54.68 55.15 50.26 55.93 50.89 61.26 52.00 66.40 52.89 57.82 55.47 52.98 48.89 68.66 57.53 76.84 63.00 79.10 59.47
Qian_nivic_task2_1 Qiannivic2026 23 62.311119555002314 ± 0.002896607555457615 71.42 61.21 47.12 49.68 67.89 58.05 62.87 55.68 86.38 70.37 75.43 62.21 77.81 59.74 54.37 51.37 59.52 50.15 75.27 59.00 75.64 57.53 70.96 53.42
Qian_nivic_task2_2 Qiannivic2026 8 64.42767291560604 ± 0.002963997006229839 72.85 62.47 49.35 50.26 69.47 58.42 66.37 59.26 87.24 76.37 73.41 58.84 77.56 60.63 56.52 52.16 62.42 50.26 73.65 57.05 74.89 57.37 76.39 60.73
Qian_nivic_task2_3 Qiannivic2026 42 61.18003358708476 ± 0.0027110369824390047 69.20 58.37 44.66 51.00 69.58 56.21 63.13 53.42 86.98 67.47 75.10 59.74 74.80 54.37 53.27 52.11 61.92 49.58 73.77 57.16 73.99 59.00 71.83 58.21
Qian_nivic_task2_4 Qiannivic2026 12 63.4466709309529 ± 0.0028881543367683406 70.62 62.16 47.35 51.26 69.69 55.84 66.76 58.37 86.96 73.74 74.20 57.53 75.48 55.37 55.76 52.16 66.82 49.74 75.00 59.42 74.29 59.37 76.66 67.79
Jeong_Medisensing_task2_1 JeongMedisensing2026 119 56.853596513643865 ± 0.002885157183710774 48.47 52.58 50.54 49.79 60.56 57.68 55.34 51.26 81.48 74.95 61.40 51.95 69.84 58.32 59.14 54.89 59.84 50.89 72.80 54.89 78.46 63.16 85.94 63.11
Jeong_Medisensing_task2_2 JeongMedisensing2026 135 55.020142655920814 ± 0.0027687158913131792 42.66 51.26 48.34 49.32 60.95 58.84 55.94 52.21 78.15 70.00 58.28 52.47 70.06 57.53 58.40 53.79 59.00 50.95 72.22 56.89 77.84 64.53 83.58 63.53
Jeong_Medisensing_task2_3 JeongMedisensing2026 125 56.139287428650995 ± 0.0027990995390076807 48.07 52.53 48.30 48.95 60.87 57.95 55.85 51.79 79.47 69.89 59.64 51.58 69.72 57.53 57.10 52.79 59.32 51.53 70.12 56.11 79.00 62.42 87.20 69.05
Jeong_Medisensing_task2_4 JeongMedisensing2026 106 57.435901782845434 ± 0.0029330489767252477 52.95 53.11 51.79 50.42 57.55 55.11 54.39 50.61 84.19 77.37 65.62 53.05 67.91 57.74 60.54 58.16 60.54 51.53 69.89 51.21 69.31 56.24 84.93 57.26
Zhou_XAUAT_task2_1 ZhouXAUAT2026 37 61.37081867585763 ± 0.0027780323220549705 65.09 56.58 53.72 52.42 58.92 54.58 66.22 57.95 80.16 66.53 85.83 74.74 89.71 81.58 63.69 59.63 63.95 52.74 70.44 54.74 66.58 55.37 61.87 52.63
Kim_CAU_task2_1 KimCAU2026 101 57.580567978413654 ± 0.0031525705411049724 59.21 60.58 45.64 47.79 59.38 60.32 55.08 52.58 82.88 63.47 65.24 55.26 67.68 53.05 60.40 56.63 63.84 53.95 73.66 59.37 82.60 63.74 76.00 60.58
Kim_CAU_task2_2 KimCAU2026 66 59.85423786944168 ± 0.003219819206690976 69.73 56.21 49.85 48.26 57.25 61.21 59.71 54.21 84.27 61.16 62.00 57.00 66.12 53.16 59.46 54.47 62.82 51.16 78.00 60.11 72.98 60.84 79.56 62.53
Kim_CAU_task2_3 KimCAU2026 86 58.48353702028144 ± 0.0030633496442016494 63.99 60.21 47.38 47.68 60.48 61.21 56.15 52.63 79.33 61.74 63.88 54.63 69.90 55.53 57.56 59.95 63.18 53.84 73.30 56.89 79.38 62.37 78.20 63.89
Kim_CAU_task2_4 KimCAU2026 81 58.92205733899948 ± 0.003054064590582812 64.81 60.68 48.21 47.79 59.42 61.42 57.00 52.00 80.74 63.32 64.76 54.84 69.70 54.89 58.36 58.63 63.58 52.74 75.14 58.32 80.56 62.68 81.62 67.47
Zeng_BUCT_task2_1 ZengBUCT2026 165 50.715623590642686 ± 0.00238905983347624 48.69 52.00 45.93 48.16 54.77 51.58 48.49 52.16 56.64 51.26 59.18 52.26 73.34 60.11 59.38 56.79 58.52 52.63 72.80 62.53 79.22 60.74 85.14 80.05
Zeng_BUCT_task2_2 ZengBUCT2026 171 49.309581912167786 ± 0.0024577082696230546 55.87 52.21 49.74 51.11 44.44 49.47 46.08 49.00 49.35 50.21 80.31 65.37 75.96 61.47 66.23 58.84 59.90 51.63 65.19 48.37 68.49 58.89 59.01 66.78
Zeng_BUCT_task2_3 ZengBUCT2026 172 49.267213448948624 ± 0.0024107082554433216 53.16 50.37 51.92 53.16 42.21 49.53 44.98 47.68 54.06 50.89 71.21 63.71 76.16 56.96 63.72 61.53 64.00 52.37 71.69 53.24 73.25 65.12 82.04 63.64
Zeng_BUCT_task2_4 ZengBUCT2026 162 51.10928407930502 ± 0.0024914923213899603 48.80 52.00 45.74 47.95 56.52 51.79 48.65 52.37 58.62 50.84 56.26 53.63 74.60 62.32 53.92 54.21 59.38 50.74 70.04 62.42 75.40 58.26 84.72 77.79
Yang_NJU_task2_1 YangNJU2026 43 61.1737276780872 ± 0.002856466985997388 71.07 57.11 45.68 48.16 57.26 51.16 72.62 60.84 85.05 76.74 0.60 0.58 0.69 0.55 0.62 0.56 0.59 0.50 0.73 0.59 0.82 0.63 0.67 0.52
Yang_NJU_task2_2 YangNJU2026 57 60.35045060661118 ± 0.002883236432286248 67.27 54.37 45.08 48.16 56.89 51.16 73.89 61.16 84.75 73.68 0.68 0.59 0.74 0.57 0.59 0.52 0.62 0.52 0.67 0.49 0.82 0.66 0.65 0.50
Kim_KATECH_task2_1 KimKATECH2026 73 59.49826595354606 ± 0.0023779583703373236 61.27 58.68 52.96 57.58 60.05 55.00 56.23 50.58 78.66 66.47 64.05 60.42 71.42 58.79 62.25 59.89 63.95 53.21 72.38 53.47 75.80 61.26 81.80 75.53
Kim_KATECH_task2_2 KimKATECH2026 80 58.985052161247694 ± 0.0023812230771927275 61.56 58.68 51.50 57.21 59.57 55.26 55.46 49.84 78.69 65.58 64.41 57.47 70.42 57.11 62.66 59.37 63.99 53.26 73.67 53.53 77.20 62.74 65.79 52.58
Kim_KATECH_task2_3 KimKATECH2026 61 60.21000436812732 ± 0.0023850928221648194 60.92 59.58 53.44 56.74 61.44 54.42 57.83 50.58 82.26 67.26 75.70 54.68 71.15 58.32 62.39 59.32 61.28 51.95 67.35 50.47 78.37 60.58 53.37 50.37
Kim_KATECH_task2_4 KimKATECH2026 77 59.3878699958256 ± 0.0023995388800558707 61.32 59.00 52.73 57.42 59.73 55.37 56.04 50.00 78.66 66.63 62.68 58.26 70.67 58.00 62.49 60.00 63.29 53.68 72.21 54.00 77.10 62.47 82.91 76.84
Guan_GISP@HEU_task2_1 GuanGISP@HEU2026 79 59.00784814563139 ± 0.0031432640274972243 60.59 61.11 55.98 53.53 62.48 59.11 49.21 48.74 77.71 68.53 60.50 54.20 60.25 56.10 60.50 58.90 59.65 50.00 71.90 51.30 77.40 56.10 70.95 50.60
Krag_AAU_task2_1 KragAAU2026 38 61.328080453648845 ± 0.0028201240454685 60.24 58.47 57.89 50.53 59.55 55.37 60.80 53.21 86.71 73.84 62.64 55.95 74.02 57.58 64.38 57.37 68.44 54.74 75.60 55.37 78.02 53.84 91.08 77.68
Krag_AAU_task2_2 KragAAU2026 104 57.47234990217671 ± 0.002723376487894248 55.89 53.76 57.37 50.55 59.80 56.71 55.92 52.08 64.92 66.16 72.25 55.92 75.86 58.13 64.62 53.82 66.33 53.58 75.51 57.50 77.55 57.18 89.94 77.00
Krag_AAU_task2_3 KragAAU2026 116 56.96096545694428 ± 0.002763420472448576 59.77 54.74 51.95 49.21 59.02 55.95 56.48 52.58 64.41 64.32 62.72 53.16 76.32 68.53 62.86 59.21 73.62 57.16 78.28 56.00 81.32 55.47 91.98 82.42
Krag_AAU_task2_4 KragAAU2026 19 62.58781647688 ± 0.0029384679344328225 64.44 58.24 62.35 54.61 58.76 54.13 61.76 54.50 81.92 74.58 67.34 52.53 79.66 68.34 64.63 57.84 70.91 55.45 80.91 71.13 80.38 63.21 89.61 81.45
Moradi_JKU_task2_1 MoradiJKU2026 52 60.79515879820278 ± 0.003065461869000823 57.63 53.21 52.47 53.95 70.82 65.53 66.08 53.53 72.00 61.16 59.68 58.58 63.02 52.89 63.36 62.21 62.84 54.74 70.02 49.11 70.62 53.47 81.30 72.42
Moradi_JKU_task2_2 MoradiJKU2026 60 60.213851357458225 ± 0.003029188148818241 55.81 53.00 51.44 54.63 70.73 65.21 65.52 53.68 70.71 61.84 60.98 58.84 63.14 53.21 63.40 62.26 62.94 54.89 69.88 49.21 70.52 53.84 80.36 71.37


Supplementary metrics (recall, precision, and F1 score)

Rank Submission Information Evaluation Dataset
Submission Code Technical
Report 		Official
Rank 		ToyDrone
(F1 score) 		ToyDrone
(Recall) 		ToyDrone
(Precision) 		ToothBrush
(F1 score) 		ToothBrush
(Recall) 		ToothBrush
(Precision) 		SewingMachine
(F1 score) 		SewingMachine
(Recall) 		SewingMachine
(Precision) 		Sander
(F1 score) 		Sander
(Recall) 		Sander
(Precision) 		BlowerDustCollector
(F1 score) 		BlowerDustCollector
(Recall) 		BlowerDustCollector
(Precision)
DCASE2026_baseline_task2_MAHALA DCASE2026baseline2026 137 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 64.66 86.00 51.81 66.67 100.00 50.00
DCASE2026_baseline_task2_MSE DCASE2026baseline2026 68 58.30 61.61 55.32 66.67 100.00 50.00 61.68 72.66 53.59 65.73 86.44 53.03 69.94 98.99 54.07
Ozeki_MELCO_task2_1 OzekiMELCO2026 121 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Ozeki_MELCO_task2_2 OzekiMELCO2026 107 49.14 38.48 67.99 70.78 86.00 60.14 64.42 76.99 55.38 64.80 78.99 54.94 59.17 67.76 52.51
Ozeki_MELCO_task2_3 OzekiMELCO2026 59 41.41 29.65 68.66 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Ozeki_MELCO_task2_4 OzekiMELCO2026 49 64.47 93.62 49.16 56.87 60.59 53.58 64.78 81.95 53.55 65.89 79.95 56.04 74.71 89.82 63.96
Qian_SJTU_task2_1 QianSJTU2026 89 63.80 63.94 63.66 51.13 51.31 50.95 50.00 50.00 50.00 59.52 59.40 59.64 66.74 67.06 66.43
Qian_SJTU_task2_2 QianSJTU2026 54 56.00 56.00 56.00 58.00 58.00 58.00 54.96 54.84 55.08 56.56 56.56 56.56 73.74 73.92 73.57
Qian_SJTU_task2_3 QianSJTU2026 91 63.08 62.98 63.18 49.60 49.68 49.52 49.00 48.82 49.18 58.58 58.58 58.58 67.56 67.83 67.30
Qian_SJTU_task2_4 QianSJTU2026 88 64.23 63.75 64.72 48.44 48.49 48.39 50.92 50.98 50.86 60.82 60.85 60.79 70.25 70.65 69.86
Qian_VUILabs_task2_1 QianVUILabs2026 55 62.00 62.00 62.00 54.75 54.84 54.66 52.98 52.83 53.13 63.91 63.75 64.07 78.54 78.91 78.16
Qian_VUILabs_task2_2 QianVUILabs2026 78 55.93 55.58 56.29 50.77 50.77 50.77 56.84 56.84 56.84 51.37 51.31 51.43 77.25 77.47 77.04
Qian_VUILabs_task2_3 QianVUILabs2026 56 61.42 61.42 61.42 48.72 48.72 48.72 51.97 51.69 52.26 59.00 58.98 59.02 73.07 73.39 72.75
Qian_VUILabs_task2_4 QianVUILabs2026 63 57.05 56.90 57.21 51.65 51.47 51.82 56.28 56.14 56.42 51.88 51.92 51.84 75.11 75.43 74.79
Zhang_XJTLU_task2_1 ZhangXJTLU2026 133 0.00 0.00 0.00 0.00 0.00 0.00 5.80 3.00 85.71 0.00 0.00 0.00 6.33 3.33 62.50
Zhang_XJTLU_task2_2 ZhangXJTLU2026 134 0.00 0.00 0.00 0.00 0.00 0.00 5.80 3.00 85.71 0.00 0.00 0.00 6.33 3.33 62.50
Zhang_XJTLU_task2_3 ZhangXJTLU2026 132 0.00 0.00 0.00 0.00 0.00 0.00 5.80 3.00 85.71 0.00 0.00 0.00 6.33 3.33 62.50
Zhang_XJTLU_task2_4 ZhangXJTLU2026 131 0.00 0.00 0.00 0.00 0.00 0.00 5.80 3.00 85.71 0.00 0.00 0.00 6.33 3.33 62.50
Chang_Surrey_task2_1 ChangSurrey2026 58 28.17 18.11 63.32 54.84 51.53 58.62 70.10 69.94 70.26 60.13 59.73 60.54 71.82 72.05 71.59
Chang_Surrey_task2_2 ChangSurrey2026 76 28.04 18.08 62.50 53.71 50.67 57.14 63.55 63.75 63.35 54.60 54.11 55.09 67.65 67.76 67.53
Chang_Surrey_task2_3 ChangSurrey2026 112 23.85 14.77 61.94 60.13 59.93 60.34 62.92 62.22 63.64 55.78 55.36 56.20 64.00 63.94 64.06
Chang_Surrey_task2_4 ChangSurrey2026 142 34.87 28.80 44.17 44.98 42.69 47.52 52.58 52.53 52.63 50.96 50.51 51.42 66.27 66.27 66.27
Zhang_SATLab_task2_1 ZhangSATLab2026 45 64.68 64.86 64.50 53.97 53.53 54.42 52.00 52.00 52.00 61.96 61.94 61.98 68.34 68.64 68.05
Zhang_SATLab_task2_2 ZhangSATLab2026 71 60.15 59.93 60.38 55.60 55.71 55.48 54.79 54.55 55.05 62.00 62.00 62.00 75.11 75.53 74.70
Zhang_SATLab_task2_3 ZhangSATLab2026 39 65.80 65.94 65.66 55.30 54.88 55.74 50.00 49.92 50.08 60.82 60.85 60.79 73.93 74.35 73.52
Zhang_SATLab_task2_4 ZhangSATLab2026 44 66.12 65.45 66.79 56.37 56.28 56.47 51.94 51.92 51.96 60.85 60.85 60.85 74.10 74.68 73.52
Fan_WISTLAB_task2_1 FanWISTLAB2026 35 62.22 62.22 62.22 54.35 54.55 54.15 55.41 54.86 55.98 56.98 56.98 56.98 66.38 66.63 66.13
Fan_WISTLAB_task2_2 FanWISTLAB2026 33 60.40 60.20 60.61 52.80 52.83 52.77 55.79 55.71 55.87 56.21 56.14 56.28 66.43 66.63 66.23
Fan_WISTLAB_task2_3 FanWISTLAB2026 34 60.34 60.20 60.47 52.48 52.53 52.43 55.83 55.71 55.95 56.28 56.14 56.42 66.43 66.63 66.23
Fan_WISTLAB_task2_4 FanWISTLAB2026 17 62.55 62.60 62.50 51.25 50.72 51.79 53.78 53.70 53.86 61.98 61.94 62.02 76.79 77.18 76.40
Jiang_AITHU_task2_1 JiangAITHU2026 98 62.00 62.00 62.00 48.94 48.49 49.40 48.67 48.82 48.52 60.88 60.85 60.91 65.16 65.45 64.86
Jiang_AITHU_task2_2 JiangAITHU2026 82 62.91 62.98 62.84 49.40 49.28 49.52 51.03 50.98 51.08 61.92 61.94 61.90 71.92 72.33 71.51
Jiang_AITHU_task2_3 JiangAITHU2026 21 64.52 64.62 64.42 55.09 54.86 55.33 56.00 56.00 56.00 58.91 58.85 58.97 71.92 72.33 71.51
Jiang_AITHU_task2_4 JiangAITHU2026 41 68.00 68.00 68.00 56.45 55.52 57.42 54.03 53.93 54.13 58.78 58.58 58.98 75.11 75.53 74.70
Zhang_THUEE_task2_1 ZhangTHUEE2026 30 59.98 57.27 62.97 54.95 54.11 55.81 55.05 54.84 55.26 58.21 57.72 58.70 78.46 78.75 78.16
Zhang_THUEE_task2_2 ZhangTHUEE2026 51 65.19 64.86 65.53 54.62 54.86 54.39 52.00 52.00 52.00 61.96 61.94 61.98 74.61 75.16 74.07
Zhang_THUEE_task2_3 ZhangTHUEE2026 24 68.08 67.94 68.22 54.35 54.55 54.15 52.98 52.98 52.98 60.85 60.85 60.85 73.39 73.92 72.87
Zhang_THUEE_task2_4 ZhangTHUEE2026 11 61.59 60.32 62.91 57.92 57.38 58.47 56.95 56.98 56.92 59.06 58.85 59.27 74.92 75.16 74.69
Huang_CQUPT_task2_1 HuangCQUPT2026 111 65.28 96.91 49.21 66.89 100.00 50.25 62.13 77.95 51.65 64.64 84.89 52.18 66.19 94.99 50.79
Huang_CQUPT_task2_2 HuangCQUPT2026 124 47.91 41.10 57.44 27.54 18.95 50.42 50.26 36.76 79.44 42.97 33.60 59.57 63.98 69.77 59.07
Huang_CQUPT_task2_3 HuangCQUPT2026 122 64.26 85.61 51.43 59.37 71.35 50.83 55.64 52.83 58.77 56.64 57.93 55.41 62.86 79.80 51.85
Huang_CQUPT_task2_4 HuangCQUPT2026 109 65.32 96.99 49.24 65.98 97.96 49.74 63.43 79.95 52.56 64.73 83.24 52.95 67.35 98.00 51.31
Xie_SHU_task2_1 XieSHU2026 75 44.62 39.60 51.10 53.95 56.14 51.93 59.64 52.15 69.63 50.63 43.64 60.30 68.50 71.94 65.37
Xie_SHU_task2_2 XieSHU2026 65 44.02 37.74 52.80 57.20 57.38 57.03 59.49 50.51 72.36 53.61 47.67 61.24 71.29 75.95 67.16
Xie_SHU_task2_3 XieSHU2026 84 40.54 32.97 52.61 54.55 51.92 57.45 56.54 50.82 63.72 52.59 46.47 60.57 68.09 64.00 72.73
Xie_SHU_task2_4 XieSHU2026 93 46.42 39.60 56.09 50.87 47.67 54.53 53.54 44.98 66.14 51.40 44.61 60.64 66.34 61.42 72.12
Moon_Independent_task2_1 MoonIndependent2026 127 52.19 48.86 56.02 47.69 44.68 51.14 56.72 57.63 55.85 59.03 58.93 59.13 55.85 56.90 54.84
Xia_NEU_task2_1 XiaNEU2026 32 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Xia_NEU_task2_2 XiaNEU2026 69 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Xia_NEU_task2_3 XiaNEU2026 67 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Xia_NEU_task2_4 XiaNEU2026 46 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
XingWu_MCPX_task2_1 XingWuMCPX2026 50 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
XingWu_MCPX_task2_2 XingWuMCPX2026 36 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
XingWu_MCPX_task2_3 XingWuMCPX2026 74 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
XingWu_MCPX_task2_4 XingWuMCPX2026 29 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Zhou_SUMERUZOO_task2_1 ZhouSUMERUZOO2026 175 0.00 0.00 0.00 47.71 48.00 47.43 28.83 19.20 57.83 29.13 19.81 55.03 15.38 9.43 41.77
Zhou_SUMERUZOO_task2_3 ZhouSUMERUZOO2026 174 7.25 3.84 64.86 5.92 3.33 26.32 4.71 2.67 20.00 10.08 5.71 42.55 27.69 18.95 51.43
Kwon_KIST_task2_1 KwonKIST2026 110 47.25 40.92 55.88 52.15 52.15 52.15 52.00 52.00 52.00 60.00 60.00 60.00 68.54 68.64 68.44
Kwon_KIST_task2_2 KwonKIST2026 114 44.91 38.75 53.41 52.69 52.80 52.59 53.95 53.93 53.97 61.03 60.98 61.08 68.54 68.64 68.44
Kwon_KIST_task2_3 KwonKIST2026 120 44.91 38.75 53.41 52.58 52.81 52.35 53.78 53.70 53.86 60.05 59.93 60.17 69.49 69.49 69.49
Kwon_KIST_task2_4 KwonKIST2026 113 47.00 40.92 55.19 51.96 52.15 51.76 53.95 53.93 53.97 60.00 60.00 60.00 69.49 69.49 69.49
Fujimura_MERL_task2_1 FujimuraMERL2026 2 54.14 44.48 69.17 57.43 59.43 55.57 58.38 50.82 68.57 58.91 54.98 63.45 71.43 100.00 55.56
Fujimura_MERL_task2_2 FujimuraMERL2026 15 44.20 30.92 77.51 48.55 46.04 51.35 61.02 50.82 76.33 49.73 39.80 66.23 71.15 85.58 60.89
Fujimura_MERL_task2_3 FujimuraMERL2026 1 57.53 47.36 73.26 64.98 71.74 59.39 59.89 51.69 71.19 56.27 46.47 71.33 75.47 100.00 60.61
Fujimura_MERL_task2_4 FujimuraMERL2026 6 49.48 37.62 72.28 53.37 52.81 53.93 60.55 50.51 75.59 59.09 50.77 70.66 74.78 95.96 61.27
Noh_CBNU_task2_1 NohCBNU2026 118 55.15 52.41 58.19 52.62 51.33 53.97 56.28 55.93 56.64 56.99 56.84 57.14 61.94 61.94 61.94
Jeong_KETI_task2_1 JeongKETI2026 130 28.04 21.12 41.71 55.62 53.33 58.11 17.09 10.00 58.82 42.44 30.97 67.42 38.57 31.11 50.72
Zarrouky_IR_task2_1 ZarroukyIR2026 140 54.03 53.93 54.13 49.20 48.82 49.59 54.15 53.53 54.78 50.17 50.51 49.83 64.39 63.75 65.03
Lei_CRRC_task2_1 LeiCRRC2026 136 66.67 100.00 50.00 66.67 100.00 50.00 65.28 96.91 49.21 66.88 97.96 50.77 66.67 100.00 50.00
Lei_CRRC_task2_2 LeiCRRC2026 128 66.67 100.00 50.00 66.67 100.00 50.00 65.49 96.91 49.46 65.45 90.99 51.11 66.67 100.00 50.00
Lei_CRRC_task2_3 LeiCRRC2026 149 66.67 100.00 50.00 66.22 98.99 49.75 66.67 100.00 50.00 65.97 94.99 50.53 69.93 93.83 55.74
Lei_CRRC_task2_4 LeiCRRC2026 108 66.67 93.62 51.76 65.86 80.99 55.50 61.05 65.79 56.95 62.99 74.88 54.36 65.10 88.90 51.36
Tsz_HFU_task2_1 TszHFU2026 126 66.67 100.00 50.00 66.67 100.00 50.00 62.40 81.80 50.44 64.14 83.95 51.90 66.67 100.00 50.00
Tsz_HFU_task2_2 TszHFU2026 156 0.00 0.00 0.00 13.22 7.50 55.56 21.40 12.86 63.83 20.00 12.00 60.00 7.19 3.75 88.24
Tsz_HFU_task2_3 TszHFU2026 103 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 40.21 28.97 65.73 31.29 20.95 61.80
Tsz_HFU_task2_4 TszHFU2026 170 0.00 0.00 0.00 17.49 10.67 48.48 6.96 4.00 26.67 0.00 0.00 0.00 15.04 8.89 48.78
Kim_LUDO_task2_1 KimLUDO2026 70 48.72 40.92 60.20 45.06 42.00 48.61 53.57 43.64 69.36 45.51 37.05 58.96 68.55 81.22 59.29
Kim_LUDO_task2_2 KimLUDO2026 64 44.33 35.36 59.42 46.04 46.04 46.05 57.44 48.00 71.49 50.79 42.42 63.29 69.23 86.44 57.73
Kim_LUDO_task2_3 KimLUDO2026 53 50.55 45.71 56.54 50.55 54.10 47.43 57.01 46.98 72.49 46.45 38.77 57.93 68.63 83.24 58.38
Kim_LUDO_task2_4 KimLUDO2026 102 45.51 34.32 67.51 44.75 39.10 52.31 57.23 47.25 72.55 44.89 34.21 65.26 61.61 56.84 67.25
Balozi_RISE_task2_1 BaloziRISE2026 155 22.66 14.69 49.45 51.79 50.72 52.91 56.63 56.14 57.12 52.21 52.08 52.35 60.01 59.93 60.09
Balozi_RISE_task2_2 BaloziRISE2026 154 18.29 11.25 48.91 52.99 52.08 53.93 54.01 53.93 54.09 52.63 52.53 52.73 62.98 62.98 62.98
Balozi_RISE_task2_3 BaloziRISE2026 166 13.50 7.67 56.12 49.00 48.82 49.18 58.83 58.58 59.08 54.87 54.84 54.90 62.99 62.98 63.00
Balozi_RISE_task2_4 BaloziRISE2026 157 22.56 14.67 48.89 55.03 54.98 55.08 56.70 56.14 57.27 52.63 52.53 52.73 58.91 58.85 58.97
Mei_FDID_task2_1 MeiFDID2026 129 53.23 52.15 54.36 52.77 52.08 53.49 58.52 58.17 58.87 53.77 52.81 54.76 63.03 62.98 63.08
Mei_FDID_task2_2 MeiFDID2026 138 52.12 50.72 53.61 47.85 47.25 48.46 53.78 53.70 53.86 57.06 56.56 57.57 63.99 63.75 64.23
Mei_FDID_task2_3 MeiFDID2026 150 42.56 36.71 50.65 50.59 50.04 51.16 54.93 54.84 55.02 55.78 55.36 56.20 59.40 59.40 59.40
Wang_WST_task2_1 WangWST2026 148 7.15 3.73 84.85 20.00 12.00 60.00 15.00 8.73 53.33 6.52 3.56 39.02 25.57 14.74 96.55
Wang_WST_task2_2 WangWST2026 173 17.80 10.87 49.01 38.62 31.06 51.06 13.39 7.50 62.50 13.48 8.40 34.15 41.48 30.88 63.17
Wang_WST_task2_3 WangWST2026 145 7.15 3.73 84.85 20.00 12.00 60.00 22.97 13.71 70.59 6.62 3.60 40.91 28.77 16.80 100.00
Jiang_KY_task2_1 JiangKY2026 167 51.21 50.51 51.94 46.81 46.81 46.81 48.00 48.00 48.00 58.71 56.28 61.35 59.76 59.67 59.85
Jiang_KY_task2_2 JiangKY2026 160 50.24 49.68 50.82 48.74 48.49 48.99 51.92 51.92 51.92 57.67 55.58 59.93 52.08 52.08 52.08
Jiang_KY_task2_3 JiangKY2026 141 54.78 53.53 56.10 50.56 50.51 50.61 47.67 47.67 47.67 60.55 58.33 62.95 59.76 58.23 61.37
Jiang_KY_task2_4 JiangKY2026 144 53.55 49.75 57.99 57.74 57.38 58.10 46.84 46.81 46.87 60.55 58.33 62.95 59.57 58.93 60.22
Glitza_IKA_task2_1 GlitzaIKA2026 153 23.49 14.51 61.54 37.21 27.43 57.83 8.92 4.80 63.16 34.30 24.83 55.47 32.05 20.74 70.53
Glitza_IKA_task2_2 GlitzaIKA2026 147 65.96 96.91 50.00 22.26 14.12 52.63 66.44 98.99 50.00 65.01 92.90 50.00 66.22 98.00 50.00
Glitza_IKA_task2_3 GlitzaIKA2026 139 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Glitza_IKA_task2_4 GlitzaIKA2026 164 66.67 100.00 50.00 11.68 6.86 39.34 66.34 88.90 52.92 64.21 91.30 49.52 66.57 82.89 55.62
Kajita_IND_task2_1 KajitaIND2026 159 0.00 0.00 0.00 7.55 4.00 66.67 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Kajita_IND_task2_2 KajitaIND2026 161 0.00 0.00 0.00 7.55 4.00 66.67 5.06 2.67 50.00 0.00 0.00 0.00 0.00 0.00 0.00
Kajita_IND_task2_3 KajitaIND2026 151 0.00 0.00 0.00 7.55 4.00 66.67 5.69 3.00 54.55 0.00 0.00 0.00 0.00 0.00 0.00
Kajita_IND_task2_4 KajitaIND2026 158 0.00 0.00 0.00 7.55 4.00 66.67 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Zhang_JAIST_task2_1 ZhangJAIST2026 47 12.96 7.00 87.50 19.66 11.67 62.50 32.21 19.20 100.00 20.92 12.31 69.57 26.09 15.00 100.00
Wang_UniS_task2_1 WangUniS2026 72 39.23 28.64 62.25 34.34 27.10 46.88 44.27 34.74 60.98 50.89 44.44 59.52 67.10 98.99 50.76
Wang_UniS_task2_2 WangUniS2026 143 46.15 38.71 57.14 32.86 25.20 47.19 51.22 42.00 65.62 51.54 42.42 65.66 82.72 94.99 73.26
Wang_UniS_task2_3 WangUniS2026 87 0.00 0.00 0.00 13.33 8.00 40.00 31.04 18.53 95.65 19.72 11.67 63.64 30.80 18.20 100.00
Wang_UniS_task2_4 WangUniS2026 92 0.00 0.00 0.00 13.33 8.00 40.00 31.04 18.53 95.65 19.72 11.67 63.64 32.21 19.20 100.00
Yang_XJU_task2_1 YangXJU2026 99 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Yang_XJU_task2_2 YangXJU2026 85 0.00 0.00 0.00 12.77 7.50 42.86 36.62 22.61 96.30 19.39 11.67 57.38 14.55 8.00 80.00
Yang_XJU_task2_3 YangXJU2026 105 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Yang_XJU_task2_4 YangXJU2026 62 43.12 37.53 50.66 54.88 51.93 58.18 60.05 59.93 60.17 58.61 57.63 59.63 78.58 78.68 78.48
SNU_task2_1 SNUtask22026 169 32.97 24.30 51.27 17.98 12.00 35.82 23.88 16.00 47.06 35.04 26.67 51.06 58.29 53.93 63.41
Huang_WHU_task2_1 HuangWHU2026 4 31.66 20.21 73.00 54.00 46.53 64.33 47.29 30.97 100.00 42.60 30.19 72.33 66.27 53.93 85.95
Huang_WHU_task2_2 HuangWHU2026 7 31.95 20.40 73.65 55.30 47.69 65.80 46.94 30.97 96.97 39.43 27.43 70.07 65.49 54.84 81.29
Huang_WHU_task2_3 HuangWHU2026 31 64.74 64.86 64.62 45.51 45.22 45.81 60.89 60.59 61.19 55.71 55.71 55.71 74.00 74.35 73.66
Huang_WHU_task2_4 HuangWHU2026 83 7.08 3.75 62.50 10.91 6.00 60.00 43.96 29.87 83.27 27.07 17.68 57.73 39.60 26.21 81.02
Morita_KM_task2_1 MoritaKM2026 117 55.54 52.66 58.76 43.60 35.10 57.54 50.84 44.61 59.10 59.03 58.93 59.13 71.45 71.89 71.01
Morita_KM_task2_2 MoritaKM2026 96 53.17 48.48 58.87 44.76 41.74 48.24 58.60 53.70 64.47 59.58 59.40 59.76 75.35 75.49 75.22
Morita_KM_task2_3 MoritaKM2026 97 41.94 35.74 50.72 42.76 34.81 55.42 56.16 52.53 60.34 53.55 51.69 55.54 68.00 80.24 59.00
Morita_KM_task2_4 MoritaKM2026 90 50.94 47.08 55.49 43.17 35.00 56.30 49.41 39.60 65.67 44.36 36.92 55.56 68.17 65.88 70.62
Wu_CUMT_task2_1 WuCUMT2026 13 51.46 50.72 52.23 53.22 53.33 53.10 67.00 66.99 67.01 58.97 58.98 58.96 76.30 76.15 76.44
Wu_CUMT_task2_2 WuCUMT2026 9 56.02 55.58 56.47 59.08 58.98 59.18 66.00 66.00 66.00 60.00 60.00 60.00 78.66 78.22 79.10
Wu_CUMT_task2_3 WuCUMT2026 5 57.26 55.73 58.87 60.59 60.59 60.59 67.94 67.94 67.94 56.98 56.98 56.98 81.80 81.67 81.94
Wu_CUMT_task2_4 WuCUMT2026 3 59.24 58.23 60.29 56.00 56.00 56.00 66.02 65.94 66.10 54.98 54.98 54.98 83.39 83.01 83.78
Yang_None_task2_1 YangNone2026 10 60.47 58.94 62.08 56.72 56.84 56.60 63.01 62.86 63.16 56.87 56.84 56.90 78.66 78.91 78.41
Yang_None_task2_2 YangNone2026 14 56.09 53.77 58.61 50.77 50.77 50.77 65.01 64.86 65.16 56.00 56.00 56.00 79.18 79.51 78.86
Yang_None_task2_3 YangNone2026 16 60.14 59.68 60.62 49.04 48.98 49.10 63.01 62.98 63.04 60.88 60.85 60.91 79.18 79.51 78.86
Yang_None_task2_4 YangNone2026 18 53.20 51.93 54.53 52.04 51.92 52.16 63.99 63.75 64.23 56.00 56.00 56.00 79.13 79.55 78.72
Zheng_HFUUAI_task2_1 ZhengHFUUAI2026 48 17.53 9.75 86.67 23.02 13.71 71.64 20.00 12.00 60.00 20.83 12.31 67.80 22.70 12.80 100.00
Zheng_HFUUAI_task2_2 ZhengHFUUAI2026 40 12.63 6.86 80.00 24.86 14.93 74.17 24.92 14.93 75.17 24.94 14.93 75.68 26.09 15.00 100.00
Zheng_HFUUAI_task2_3 ZhengHFUUAI2026 25 17.82 9.88 90.32 19.34 11.67 56.45 28.24 16.94 84.71 23.33 14.00 70.00 13.33 7.16 97.14
Zheng_HFUUAI_task2_4 ZhengHFUUAI2026 123 66.67 100.00 50.00 66.67 100.00 50.00 63.73 86.71 50.38 66.67 100.00 50.00 66.67 100.00 50.00
Zhou_HFUUDS_task2_1 ZhouHFUUDS2026 28 17.82 9.88 90.32 20.00 12.00 60.00 27.53 16.47 83.83 24.97 14.93 76.19 13.33 7.16 97.14
Zhou_HFUUDS_task2_2 ZhouHFUUDS2026 22 18.07 10.00 93.75 20.92 12.31 69.57 28.26 16.94 85.21 21.46 12.92 63.16 18.51 10.20 100.00
Zhou_HFUUDS_task2_3 ZhouHFUUDS2026 26 17.82 9.88 90.32 20.00 12.00 60.00 28.24 16.94 84.71 24.97 14.93 76.19 13.33 7.16 97.14
Zhou_HFUUDS_task2_4 ZhouHFUUDS2026 20 18.07 10.00 93.75 20.92 12.31 69.57 28.26 16.94 85.21 21.46 12.92 63.16 22.70 12.80 100.00
Zhu_FDA_task2_1 ZhuFDA2026 95 24.03 14.32 74.73 18.65 11.08 59.02 37.79 24.64 81.05 44.38 33.88 64.29 69.31 69.09 69.54
Zhu_FDA_task2_2 ZhuFDA2026 115 19.27 11.05 75.27 21.98 13.33 62.50 38.24 25.38 77.46 44.10 33.88 63.16 68.08 68.29 67.88
Zhu_FDA_task2_3 ZhuFDA2026 100 19.30 11.05 76.09 11.11 6.00 75.00 36.70 24.00 77.92 33.65 23.33 60.34 65.10 57.72 74.63
Zhu_FDA_task2_4 ZhuFDA2026 94 24.32 14.44 77.08 25.85 16.47 60.09 42.83 29.87 75.68 48.27 38.97 63.39 67.04 68.57 65.57
Huang_QWS_task2_1 HuangQWS2026 27 26.75 16.15 77.78 26.13 17.68 50.00 49.32 34.97 83.61 45.23 32.94 72.16 57.53 42.00 91.30
Wang_Liu_SuzhouDongyuan_task2_1 WangLiu2026 146 57.00 59.53 54.67 47.56 44.44 51.15 56.76 60.98 53.08 53.32 54.04 52.63 66.89 100.00 50.25
Wang_Liu_SuzhouDongyuan_task2_2 WangLiu2026 163 52.04 50.75 53.41 42.18 39.90 44.73 53.46 58.00 49.57 47.55 45.33 50.00 67.34 100.00 50.76
Wang_Liu_SuzhouDongyuan_task2_3 WangLiu2026 168 50.52 48.48 52.74 42.16 39.90 44.68 52.81 56.98 49.21 47.72 45.33 50.37 65.53 96.00 49.74
Wang_Liu_SuzhouDongyuan_task2_4 WangLiu2026 152 55.36 57.67 53.22 43.15 38.40 49.23 56.89 61.94 52.60 51.57 50.37 52.84 64.35 92.99 49.20
Qian_nivic_task2_1 Qiannivic2026 23 62.99 61.54 64.52 45.51 45.22 45.81 66.08 65.94 66.22 58.38 58.17 58.59 74.84 75.16 74.53
Qian_nivic_task2_2 Qiannivic2026 8 66.71 65.88 67.55 47.88 47.92 47.84 66.00 66.00 66.00 61.44 60.97 61.93 73.07 73.39 72.75
Qian_nivic_task2_3 Qiannivic2026 42 64.29 63.82 64.76 45.75 45.22 46.30 64.00 64.00 64.00 58.88 58.85 58.91 78.71 78.89 78.53
Qian_nivic_task2_4 Qiannivic2026 12 62.65 62.40 62.90 46.92 46.47 47.38 65.00 64.98 65.02 61.78 61.42 62.14 76.94 77.18 76.71
Jeong_Medisensing_task2_1 JeongMedisensing2026 119 0.00 0.00 0.00 27.66 18.20 57.59 38.54 26.71 69.13 35.50 24.00 68.18 57.93 42.00 93.33
Jeong_Medisensing_task2_2 JeongMedisensing2026 135 7.34 3.86 73.68 23.53 16.00 44.44 45.16 32.24 75.35 37.91 26.71 65.27 60.42 44.98 92.00
Jeong_Medisensing_task2_3 JeongMedisensing2026 125 7.21 3.79 73.47 29.67 20.36 54.63 41.26 28.80 72.73 34.68 24.00 62.50 58.52 42.98 91.67
Jeong_Medisensing_task2_4 JeongMedisensing2026 106 13.55 7.41 79.37 21.51 13.33 55.56 34.64 23.04 69.73 30.84 20.57 61.54 50.22 33.53 100.00
Zhou_XAUAT_task2_1 ZhouXAUAT2026 37 59.83 59.93 59.73 56.85 57.63 56.09 53.18 51.93 54.50 60.88 60.39 61.38 74.11 73.95 74.27
Kim_CAU_task2_1 KimCAU2026 101 38.20 26.98 65.40 6.20 3.56 24.24 45.57 32.73 75.00 33.47 23.04 61.15 76.98 82.41 72.23
Kim_CAU_task2_2 KimCAU2026 66 33.52 21.76 72.92 0.00 0.00 0.00 40.11 28.80 66.06 42.41 30.71 68.49 77.82 78.38 77.26
Kim_CAU_task2_3 KimCAU2026 86 43.09 32.16 65.29 15.69 9.60 42.86 44.00 32.24 69.27 39.77 27.87 69.44 76.02 89.60 66.01
Kim_CAU_task2_4 KimCAU2026 81 39.94 29.52 61.71 19.05 12.00 46.15 44.78 33.53 67.38 38.24 26.71 67.27 76.28 86.44 68.25
Zeng_BUCT_task2_1 ZengBUCT2026 165 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Zeng_BUCT_task2_2 ZengBUCT2026 171 43.18 31.26 69.80 4.97 2.67 36.36 17.49 10.67 48.48 66.67 100.00 50.00 74.55 93.62 61.94
Zeng_BUCT_task2_3 ZengBUCT2026 172 43.18 31.26 69.80 4.97 2.67 36.36 17.49 10.67 48.48 66.67 100.00 50.00 74.55 93.62 61.94
Zeng_BUCT_task2_4 ZengBUCT2026 162 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00 66.67 100.00 50.00
Yang_NJU_task2_1 YangNJU2026 43 35.69 23.44 74.78 10.21 6.00 34.29 0.00 0.00 0.00 7.69 4.00 100.00 82.66 95.83 72.67
Yang_NJU_task2_2 YangNJU2026 57 59.19 61.11 57.38 9.78 5.71 33.90 0.00 0.00 0.00 5.80 3.00 85.71 83.65 93.62 75.60
Kim_KATECH_task2_1 KimKATECH2026 73 66.67 100.00 50.00 66.67 100.00 50.00 58.87 65.94 53.18 63.61 83.81 51.25 65.05 84.05 53.05
Kim_KATECH_task2_2 KimKATECH2026 80 66.67 100.00 50.00 66.67 100.00 50.00 60.16 73.14 51.10 64.68 85.95 51.85 67.63 92.73 53.23
Kim_KATECH_task2_3 KimKATECH2026 61 66.67 100.00 50.00 66.67 100.00 50.00 58.87 65.94 53.18 63.61 83.81 51.25 64.82 84.05 52.75
Kim_KATECH_task2_4 KimKATECH2026 77 66.67 100.00 50.00 66.67 100.00 50.00 59.80 68.64 52.98 63.05 82.70 50.94 65.15 86.86 52.12
Guan_GISP@HEU_task2_1 GuanGISP@HEU2026 79 47.66 37.53 65.28 51.65 56.14 47.83 51.04 38.97 73.93 44.92 37.74 55.46 70.31 82.02 61.52
Krag_AAU_task2_1 KragAAU2026 38 51.45 44.75 60.52 62.01 68.62 56.56 61.28 80.89 49.32 64.33 82.89 52.56 73.31 92.47 60.73
Krag_AAU_task2_2 KragAAU2026 104 55.61 55.55 55.67 60.51 66.53 55.48 60.08 69.49 52.92 57.84 62.86 53.57 61.25 68.87 55.15
Krag_AAU_task2_3 KragAAU2026 116 66.42 81.56 56.01 58.67 63.03 54.88 65.90 92.90 51.06 58.51 68.87 50.87 62.14 72.66 54.28
Krag_AAU_task2_4 KragAAU2026 19 61.56 63.77 59.49 61.86 70.00 55.41 60.66 70.99 52.95 62.39 71.94 55.07 74.29 78.75 70.31
Moradi_JKU_task2_1 MoradiJKU2026 52 66.44 95.96 50.81 64.66 92.73 49.63 68.55 88.72 55.86 69.10 92.90 55.00 67.13 95.96 51.63
Moradi_JKU_task2_2 MoradiJKU2026 60 65.47 93.83 50.28 62.60 88.45 48.44 68.66 88.90 55.92 68.97 90.00 55.90 66.89 94.91 51.65


Domain-wise performance

Rank Submission Information Ranking Eveluation Dataset in Source Domain Eveluation Dataset in Target Domain
Submission Code Technical
Report 		Official
Rank 		Official
Score 		Harmonic mean
(AUC, source) 		ToyDrone
(AUC, source) 		ToyDrone
(pAUC, source) 		ToothBrush
(AUC, source) 		ToothBrush
(pAUC, source) 		SewingMachine
(AUC, source) 		SewingMachine
(pAUC, source) 		Sander
(AUC, source) 		Sander
(pAUC, source) 		BlowerDustCollector
(AUC, source) 		BlowerDustCollector
(pAUC, source) 		Harmonic mean
(AUC, target) 		ToyDrone
(AUC, target) 		ToyDrone
(pAUC, target) 		ToothBrush
(AUC, target) 		ToothBrush
(pAUC, target) 		SewingMachine
(AUC, target) 		SewingMachine
(pAUC, target) 		Sander
(AUC, target) 		Sander
(pAUC, target) 		BlowerDustCollector
(AUC, target) 		BlowerDustCollector
(pAUC, target)
DCASE2026_baseline_task2_MAHALA DCASE2026baseline2026 137 54.763 60.86 72.76 59.26 43.98 52.79 60.34 50.58 69.80 49.47 67.66 63.89 49.85 54.28 59.26 39.40 52.79 43.76 50.58 45.10 49.47 87.20 63.89
DCASE2026_baseline_task2_MSE DCASE2026baseline2026 68 59.803 66.06 71.26 55.11 59.02 59.79 64.84 58.11 67.14 50.89 69.48 61.79 57.32 46.80 55.11 64.34 59.79 61.42 58.11 44.02 50.89 88.38 61.79
Ozeki_MELCO_task2_1 OzekiMELCO2026 121 56.806 58.86 44.68 56.47 65.18 52.89 70.50 57.79 66.86 56.63 55.30 49.95 57.14 76.86 56.47 46.14 52.89 67.66 57.79 54.12 56.63 51.10 49.95
Ozeki_MELCO_task2_2 OzekiMELCO2026 107 57.389 57.26 39.08 54.32 65.54 54.00 65.52 53.16 73.08 55.00 57.06 54.63 61.10 83.08 54.32 67.38 54.00 59.36 53.16 50.34 55.00 54.82 54.63
Ozeki_MELCO_task2_3 OzekiMELCO2026 59 60.220 71.27 76.26 60.05 60.78 55.47 69.06 52.95 81.14 50.53 72.54 66.37 55.24 54.94 60.05 46.36 55.47 49.68 52.95 49.98 50.53 94.36 66.37
Ozeki_MELCO_task2_4 OzekiMELCO2026 49 60.968 62.05 49.10 55.21 58.40 54.84 67.42 57.42 70.14 53.11 71.44 58.11 66.09 70.78 55.21 59.12 54.84 63.38 57.42 58.94 53.11 84.30 58.11
Qian_SJTU_task2_1 QianSJTU2026 89 58.219 57.86 59.34 56.68 46.86 49.63 58.66 53.89 68.32 55.47 60.48 53.32 64.01 77.46 56.68 61.40 49.63 45.60 53.89 64.04 55.47 87.98 53.32
Qian_SJTU_task2_2 QianSJTU2026 54 60.685 59.69 54.96 53.68 52.42 52.37 56.96 57.89 64.70 54.58 74.18 68.95 66.15 63.50 53.68 65.10 52.37 61.90 57.89 60.82 54.58 84.14 68.95
Qian_SJTU_task2_3 QianSJTU2026 91 58.058 57.69 60.70 60.74 44.96 50.05 55.70 51.95 68.78 54.37 64.68 57.42 62.35 81.68 60.74 57.20 50.05 41.62 51.95 65.70 54.37 89.14 57.42
Qian_SJTU_task2_4 QianSJTU2026 88 58.405 57.54 61.04 62.32 44.58 49.47 54.22 50.79 69.14 57.11 65.92 61.63 62.30 81.84 62.32 54.90 49.47 43.10 50.79 64.12 57.11 90.74 61.63
Qian_VUILabs_task2_1 QianVUILabs2026 55 60.612 60.52 57.60 59.84 47.10 50.32 61.26 54.47 65.84 50.95 79.88 71.05 65.61 76.72 59.84 63.36 50.32 50.18 54.47 60.54 50.95 91.38 71.05
Qian_VUILabs_task2_2 QianVUILabs2026 78 59.360 58.72 59.84 53.37 48.08 49.84 60.52 56.95 52.50 51.32 82.90 81.68 62.92 61.54 53.37 55.20 49.84 63.46 56.95 54.08 51.32 92.20 81.68
Qian_VUILabs_task2_3 QianVUILabs2026 56 60.393 61.01 61.82 57.58 48.34 50.11 61.84 57.05 64.60 52.58 74.42 66.00 64.60 72.38 57.58 56.96 50.11 53.14 57.05 60.50 52.58 93.64 66.00
Qian_VUILabs_task2_4 QianVUILabs2026 63 60.146 59.68 61.14 54.26 50.84 50.53 61.06 58.32 52.70 51.21 80.66 79.05 64.04 63.26 54.26 55.90 50.53 63.74 58.32 56.86 51.21 90.08 79.05
Zhang_XJTLU_task2_1 ZhangXJTLU2026 133 55.087 52.50 42.14 52.84 49.98 49.05 60.04 56.05 55.30 51.68 59.64 54.16 60.91 74.80 52.84 59.64 49.05 61.90 56.05 54.80 51.68 56.98 54.16
Zhang_XJTLU_task2_2 ZhangXJTLU2026 134 55.084 52.49 42.10 52.84 50.00 49.05 60.04 56.05 55.32 51.68 59.62 54.16 60.91 74.84 52.84 59.62 49.05 61.90 56.05 54.80 51.68 56.98 54.16
Zhang_XJTLU_task2_3 ZhangXJTLU2026 132 55.137 52.88 42.84 53.11 50.20 49.05 60.48 55.63 54.90 51.26 60.46 54.53 60.64 73.68 53.11 59.56 49.05 61.96 55.63 54.08 51.26 57.26 54.53
Zhang_XJTLU_task2_4 ZhangXJTLU2026 131 55.176 52.76 42.56 53.11 50.12 49.21 60.36 55.84 55.12 51.37 60.18 54.21 60.89 74.22 53.11 59.78 49.21 61.96 55.84 54.68 51.37 57.20 54.21
Chang_Surrey_task2_1 ChangSurrey2026 58 60.322 74.76 77.84 52.63 69.12 56.11 79.22 55.42 73.90 56.95 74.56 71.89 52.38 30.34 52.63 54.90 56.11 65.48 55.42 56.74 56.95 87.90 71.89
Chang_Surrey_task2_2 ChangSurrey2026 76 59.403 72.42 78.40 55.21 69.82 51.16 77.46 58.58 69.16 55.16 68.52 65.79 52.40 35.64 55.21 49.14 51.16 63.44 58.58 53.76 55.16 79.12 65.79
Chang_Surrey_task2_3 ChangSurrey2026 112 57.128 71.33 77.58 52.74 73.44 49.79 73.16 60.84 67.20 56.58 66.48 57.68 49.03 29.86 52.74 49.66 49.79 63.60 60.84 54.62 56.58 69.84 57.68
Chang_Surrey_task2_4 ChangSurrey2026 142 53.429 60.81 67.48 53.89 56.70 55.89 54.82 49.84 57.20 49.05 71.20 65.26 47.04 29.90 53.89 55.96 55.89 53.14 49.84 43.52 49.05 75.80 65.26
Zhang_SATLab_task2_1 ZhangSATLab2026 45 61.103 62.48 64.20 55.16 59.96 53.32 55.78 52.37 67.64 56.84 66.38 60.26 66.41 77.42 55.16 63.60 53.32 50.00 52.37 66.86 56.84 85.52 60.26
Zhang_SATLab_task2_2 ZhangSATLab2026 71 59.627 57.23 55.94 60.63 48.76 50.11 49.64 52.11 65.28 53.11 73.98 64.79 67.30 79.96 60.63 61.44 50.11 54.06 52.11 61.12 53.11 93.90 64.79
Zhang_SATLab_task2_3 ZhangSATLab2026 39 61.249 61.33 66.16 60.95 56.10 52.79 52.74 53.00 64.22 56.37 71.14 63.16 66.16 82.62 60.95 61.32 52.79 47.04 53.00 67.22 56.37 90.64 63.16
Zhang_SATLab_task2_4 ZhangSATLab2026 44 61.143 60.29 61.58 60.58 52.96 52.58 53.44 53.42 64.48 53.42 73.62 64.74 67.58 80.60 60.58 63.20 52.58 51.96 53.42 62.74 53.42 94.56 64.74
Fan_WISTLAB_task2_1 FanWISTLAB2026 35 61.597 61.30 63.48 62.32 54.64 52.47 61.06 62.26 59.84 52.26 69.32 66.11 65.34 71.72 62.32 59.30 52.47 61.88 62.26 58.40 52.26 80.44 66.11
Fan_WISTLAB_task2_2 FanWISTLAB2026 33 61.653 61.53 64.00 61.47 54.44 52.68 62.18 60.89 59.44 52.21 69.60 63.95 66.16 70.60 61.47 61.60 52.68 61.40 60.89 59.98 52.21 81.88 63.95
Fan_WISTLAB_task2_3 FanWISTLAB2026 34 61.616 61.42 64.00 61.79 54.10 52.74 62.12 60.63 59.56 52.11 69.38 64.00 66.16 70.74 61.79 61.88 52.74 61.50 60.63 59.76 52.11 81.40 64.00
Fan_WISTLAB_task2_4 FanWISTLAB2026 17 62.994 63.47 63.68 61.26 58.12 53.63 57.14 57.42 65.26 53.84 76.66 70.58 67.37 73.58 61.26 60.62 53.63 56.32 57.42 65.04 53.84 90.94 70.58
Jiang_AITHU_task2_1 JiangAITHU2026 98 57.694 55.85 62.04 61.58 42.04 50.16 55.98 52.79 66.52 55.95 59.82 59.00 62.09 77.76 61.58 58.12 50.16 42.84 52.79 65.84 55.95 83.78 59.00
Jiang_AITHU_task2_2 JiangAITHU2026 82 58.900 58.46 62.14 62.11 45.08 50.21 55.36 51.63 68.36 54.47 68.70 64.53 62.57 81.04 62.11 54.86 50.21 43.12 51.63 66.04 54.47 90.88 64.53
Jiang_AITHU_task2_3 JiangAITHU2026 21 62.363 63.07 67.86 62.16 56.32 52.95 58.14 56.21 63.02 53.47 72.90 68.32 66.54 78.04 62.16 59.72 52.95 53.40 56.21 64.54 53.47 87.98 68.32
Jiang_AITHU_task2_4 JiangAITHU2026 41 61.191 62.26 67.50 60.32 55.92 51.58 55.04 52.42 63.38 56.16 73.18 67.84 64.74 80.08 60.32 56.70 51.58 47.68 52.42 66.68 56.16 89.76 67.84
Zhang_THUEE_task2_1 ZhangTHUEE2026 30 61.870 71.48 73.16 59.47 68.56 52.05 63.26 55.47 76.28 54.26 78.24 68.79 58.50 62.86 59.47 46.74 52.05 54.26 55.47 51.74 54.26 96.04 68.79
Zhang_THUEE_task2_2 ZhangTHUEE2026 51 60.821 59.84 62.88 60.89 51.54 51.89 52.22 52.53 64.82 53.89 73.10 63.79 67.48 81.24 60.89 62.02 51.89 51.90 52.53 63.34 53.89 94.24 63.79
Zhang_THUEE_task2_3 ZhangTHUEE2026 24 62.121 61.15 65.36 62.74 52.34 53.21 54.16 55.58 64.54 53.00 74.60 67.21 68.28 81.14 62.74 62.86 53.21 53.96 55.58 63.22 53.00 93.98 67.21
Zhang_THUEE_task2_4 ZhangTHUEE2026 11 63.710 68.85 71.34 62.37 64.16 54.89 62.86 60.05 72.00 52.79 75.62 68.26 63.84 68.68 62.37 55.76 54.89 57.30 60.05 57.00 52.79 92.32 68.26
Huang_CQUPT_task2_1 HuangCQUPT2026 111 57.224 68.34 78.50 54.89 63.74 50.26 65.46 65.95 67.14 50.84 68.66 66.84 49.47 42.52 54.89 37.70 50.26 64.92 65.95 47.02 50.84 69.70 66.84
Huang_CQUPT_task2_2 HuangCQUPT2026 124 56.316 67.71 81.02 56.26 65.06 49.89 64.02 64.89 65.16 51.26 65.96 66.05 47.78 37.14 56.26 36.42 49.89 65.08 64.89 49.10 51.26 68.84 66.05
Huang_CQUPT_task2_3 HuangCQUPT2026 122 56.803 68.07 80.02 55.63 64.50 50.21 64.56 66.58 66.36 50.95 67.16 67.05 48.46 38.94 55.63 37.26 50.21 65.36 66.58 47.80 50.95 69.30 67.05
Huang_CQUPT_task2_4 HuangCQUPT2026 109 57.346 68.51 75.34 56.53 64.10 49.95 66.00 62.63 67.36 50.68 70.84 68.53 49.71 47.42 56.53 34.84 49.95 64.12 62.63 46.06 50.68 74.16 68.53
Xie_SHU_task2_1 XieSHU2026 75 59.420 59.79 59.54 50.79 45.16 52.21 68.90 65.47 66.22 61.26 66.36 57.58 61.72 42.36 50.79 70.22 52.21 73.80 65.47 60.86 61.26 75.86 57.58
Xie_SHU_task2_2 XieSHU2026 65 60.035 60.58 59.30 51.26 47.70 51.89 66.64 61.68 65.86 61.84 68.86 58.16 63.32 45.04 51.26 70.10 51.89 72.64 61.68 62.82 61.84 78.08 58.16
Xie_SHU_task2_3 XieSHU2026 84 58.719 60.02 58.70 48.95 50.12 54.05 65.08 62.00 64.90 61.00 64.36 57.79 59.95 43.12 48.95 64.50 54.05 66.72 62.00 61.22 61.00 74.74 57.79
Xie_SHU_task2_4 XieSHU2026 93 57.851 59.03 58.82 48.53 50.52 53.47 61.38 59.32 63.94 61.16 62.62 56.42 59.28 43.56 48.53 62.76 53.47 64.50 59.32 60.86 61.16 73.94 56.42
Moon_Independent_task2_1 MoonIndependent2026 127 55.609 64.64 66.98 55.00 62.70 54.26 68.34 60.00 70.36 50.79 56.72 49.84 50.32 46.20 55.00 44.96 54.26 57.28 60.00 46.56 50.79 60.48 49.84
Xia_NEU_task2_1 XiaNEU2026 32 61.721 68.70 65.96 53.84 52.12 50.68 71.14 69.16 83.52 51.47 80.60 77.21 58.75 62.02 53.84 48.10 50.68 65.76 69.16 44.72 51.47 94.16 77.21
Xia_NEU_task2_2 XiaNEU2026 69 59.732 67.63 72.60 59.11 63.32 51.26 71.80 70.00 64.34 50.37 67.14 60.89 55.43 45.54 59.11 49.14 51.26 69.68 70.00 46.96 50.37 81.64 60.89
Xia_NEU_task2_3 XiaNEU2026 67 59.830 67.88 71.82 59.00 59.72 51.53 73.74 68.47 64.38 50.47 71.96 66.53 54.80 45.06 59.00 46.70 51.53 68.96 68.47 46.78 50.47 85.02 66.53
Xia_NEU_task2_4 XiaNEU2026 46 61.069 68.55 74.66 59.47 60.76 52.95 73.14 67.26 64.22 50.16 72.26 65.21 57.57 50.48 59.47 54.42 52.95 68.46 67.26 45.50 50.16 82.78 65.21
XingWu_MCPX_task2_1 XingWuMCPX2026 50 60.954 67.43 77.32 56.63 52.44 49.89 70.02 61.21 71.10 52.32 72.46 68.26 59.43 51.88 56.63 54.72 49.89 61.96 61.21 51.54 52.32 90.54 68.26
XingWu_MCPX_task2_2 XingWuMCPX2026 36 61.534 68.27 77.88 58.26 53.16 50.21 71.64 62.16 72.20 52.16 72.58 69.21 59.71 52.26 58.26 55.92 50.21 61.58 62.16 51.40 52.16 90.68 69.21
XingWu_MCPX_task2_3 XingWuMCPX2026 74 59.478 64.04 76.84 54.47 48.50 51.95 61.08 59.37 68.76 48.84 73.92 70.89 58.76 49.48 54.47 58.64 51.95 60.36 59.37 50.56 48.84 87.14 70.89
XingWu_MCPX_task2_4 XingWuMCPX2026 29 61.875 69.12 78.16 57.84 57.12 51.37 71.08 61.47 71.34 51.84 71.70 69.05 60.03 53.18 57.84 57.64 51.37 60.92 61.47 51.22 51.84 89.24 69.05
Zhou_SUMERUZOO_task2_1 ZhouSUMERUZOO2026 175 48.363 57.29 73.56 52.16 62.32 50.32 51.30 51.11 52.24 51.68 52.62 52.05 39.78 22.50 52.16 39.62 50.32 54.30 51.11 54.12 51.68 52.30 52.05
Zhou_SUMERUZOO_task2_3 ZhouSUMERUZOO2026 174 48.437 55.30 73.98 51.00 49.52 49.11 55.50 47.58 52.99 50.11 50.46 51.53 42.05 27.82 51.00 48.31 49.11 45.05 47.58 47.34 50.11 52.79 51.53
Kwon_KIST_task2_1 KwonKIST2026 110 57.337 51.42 33.40 50.11 45.08 52.89 57.60 62.00 67.58 53.05 77.18 72.58 65.14 73.98 50.11 61.66 52.89 57.60 62.00 56.72 53.05 83.14 72.58
Kwon_KIST_task2_2 KwonKIST2026 114 57.088 51.32 33.10 49.89 45.54 52.79 57.80 62.53 67.62 53.58 75.92 70.79 64.47 72.90 49.89 61.98 52.79 56.52 62.53 56.20 53.58 81.88 70.79
Kwon_KIST_task2_3 KwonKIST2026 120 56.810 50.82 32.38 49.84 45.32 52.84 57.66 62.37 67.44 53.58 75.38 70.37 64.30 72.30 49.84 62.46 52.84 56.40 62.37 56.08 53.58 80.96 70.37
Kwon_KIST_task2_4 KwonKIST2026 113 57.111 51.41 33.32 49.84 45.56 52.84 57.80 62.58 67.48 53.63 75.86 70.68 64.42 72.72 49.84 62.28 52.84 56.62 62.58 55.96 53.63 81.42 70.68
Fujimura_MERL_task2_1 FujimuraMERL2026 2 66.004 71.72 72.38 63.21 63.54 59.26 71.70 65.58 75.12 54.05 77.48 73.53 64.52 64.78 63.21 61.40 59.26 64.38 65.58 50.76 54.05 94.88 73.53
Fujimura_MERL_task2_2 FujimuraMERL2026 15 63.217 66.95 61.58 59.68 62.30 51.00 68.80 63.74 68.66 53.63 75.20 74.11 63.78 74.42 59.68 47.54 51.00 62.24 63.74 61.14 53.63 86.94 74.11
Fujimura_MERL_task2_3 FujimuraMERL2026 1 70.241 73.81 72.52 63.53 69.30 61.74 74.92 68.79 72.20 57.37 81.12 75.74 72.79 71.08 63.53 67.44 61.74 68.28 68.79 67.04 57.37 97.70 75.74
Fujimura_MERL_task2_4 FujimuraMERL2026 6 64.997 69.80 66.60 61.89 60.72 50.26 73.06 60.47 71.50 60.05 80.18 79.00 64.72 70.14 61.89 46.28 50.26 63.86 60.47 65.00 60.05 96.68 79.00
Noh_CBNU_task2_1 NohCBNU2026 118 56.888 63.08 68.18 54.47 59.87 53.53 64.09 57.21 59.42 54.32 64.70 51.37 54.34 50.48 54.47 46.48 53.53 59.75 57.21 52.64 54.32 66.84 51.37
Jeong_KETI_task2_1 JeongKETI2026 130 55.253 57.82 65.56 55.05 57.98 51.00 47.18 50.58 59.12 56.05 63.06 59.74 53.83 42.28 55.05 58.50 51.00 52.08 50.58 57.40 56.05 64.44 59.74
Zarrouky_IR_task2_1 ZarroukyIR2026 140 53.833 60.80 64.03 51.68 52.54 50.13 57.07 51.34 63.79 48.37 69.48 57.26 50.26 54.73 51.68 45.55 50.13 51.84 51.34 38.69 48.37 70.81 57.26
Lei_CRRC_task2_1 LeiCRRC2026 136 54.903 61.01 71.96 58.58 46.68 53.58 58.92 50.89 67.56 50.11 67.32 63.89 49.87 51.20 58.58 40.48 53.58 45.26 50.89 44.92 50.11 85.64 63.89
Lei_CRRC_task2_2 LeiCRRC2026 128 55.445 61.94 73.14 59.47 47.52 52.74 61.04 52.00 67.74 50.32 67.28 64.11 50.33 48.76 59.47 40.88 52.74 48.16 52.00 45.54 50.32 85.90 64.11
Lei_CRRC_task2_3 LeiCRRC2026 149 52.789 60.98 69.66 57.84 46.76 49.21 56.86 49.95 69.08 50.47 70.46 65.63 45.67 46.16 57.84 31.88 49.21 44.60 49.95 44.32 50.47 87.26 65.63
Lei_CRRC_task2_4 LeiCRRC2026 108 57.382 69.05 75.16 50.89 66.28 51.58 66.36 62.42 67.60 54.68 70.66 63.37 50.06 31.24 50.89 48.80 51.58 66.18 62.42 53.16 54.68 74.34 63.37
Tsz_HFU_task2_1 TszHFU2026 126 55.950 63.42 72.50 58.63 50.66 52.74 59.12 50.74 70.50 49.84 70.36 64.47 51.05 51.08 58.63 44.30 52.74 44.62 50.74 46.02 49.84 85.76 64.47
Tsz_HFU_task2_2 TszHFU2026 156 51.253 60.20 75.42 52.13 61.86 50.13 55.03 52.05 57.75 51.66 55.12 54.16 44.07 26.79 52.13 50.17 50.13 57.69 52.05 46.73 51.66 57.30 54.16
Tsz_HFU_task2_3 TszHFU2026 103 57.516 57.49 58.14 50.79 53.88 53.26 57.02 55.58 56.98 53.84 62.00 50.79 63.21 64.64 50.79 67.68 53.26 64.88 55.58 58.62 53.84 61.04 50.79
Tsz_HFU_task2_4 TszHFU2026 170 49.818 48.64 54.04 51.95 42.92 49.53 43.92 47.89 53.08 55.11 51.60 50.63 49.96 54.36 51.95 42.54 49.53 39.82 47.89 64.18 55.11 57.18 50.63
Kim_LUDO_task2_1 KimLUDO2026 70 59.697 67.14 76.68 63.47 56.66 52.79 69.06 62.89 65.96 51.26 70.70 71.11 53.98 46.12 63.47 47.26 52.79 60.88 62.89 46.58 51.26 84.10 71.11
Kim_LUDO_task2_2 KimLUDO2026 64 60.109 68.64 78.32 62.74 58.52 52.63 69.26 63.68 67.60 52.00 72.68 73.26 53.66 45.28 62.74 46.36 52.63 59.54 63.68 47.62 52.00 85.28 73.26
Kim_LUDO_task2_3 KimLUDO2026 53 60.726 66.67 76.90 63.47 58.10 54.58 66.64 63.42 62.80 51.68 72.20 73.00 56.10 46.64 63.47 53.64 54.58 62.24 63.42 46.26 51.68 88.06 73.00
Kim_LUDO_task2_4 KimLUDO2026 102 57.553 68.38 75.48 52.37 60.96 49.21 71.48 63.37 67.66 51.74 68.02 64.11 51.34 40.48 52.37 45.20 49.21 62.44 63.37 47.16 51.74 75.00 64.11
Balozi_RISE_task2_1 BaloziRISE2026 155 51.267 59.76 73.84 50.47 61.11 48.53 55.42 59.42 52.40 49.68 60.10 55.84 44.00 24.62 50.47 43.44 48.53 64.94 59.42 52.81 49.68 63.86 55.84
Balozi_RISE_task2_2 BaloziRISE2026 154 51.298 60.64 72.94 49.74 60.73 48.95 57.34 60.05 52.74 50.68 63.01 54.58 43.58 22.99 49.74 45.56 48.95 64.26 60.05 53.76 50.68 66.16 54.58
Balozi_RISE_task2_3 BaloziRISE2026 166 50.695 59.50 72.44 49.74 57.80 48.89 56.12 60.37 52.74 49.21 61.92 54.95 43.03 22.18 49.74 45.78 48.89 64.62 60.37 54.19 49.21 65.15 54.95
Balozi_RISE_task2_4 BaloziRISE2026 157 51.248 59.41 73.25 51.11 59.38 48.53 55.82 59.42 51.67 49.68 60.98 55.58 44.09 24.48 51.11 43.84 48.53 66.24 59.42 52.52 49.68 64.04 55.58
Mei_FDID_task2_1 MeiFDID2026 129 55.390 64.13 66.12 50.68 59.12 50.53 69.86 58.89 66.44 49.53 60.36 56.84 50.74 47.18 50.68 42.04 50.53 62.78 58.89 43.72 49.53 67.72 56.84
Mei_FDID_task2_2 MeiFDID2026 138 54.515 63.46 64.76 50.74 59.62 50.58 61.36 56.05 71.36 49.53 61.48 53.11 49.99 45.20 50.74 40.10 50.58 56.86 56.05 47.70 49.53 69.42 53.11
Mei_FDID_task2_3 MeiFDID2026 150 52.526 61.91 73.10 50.79 59.36 50.79 57.98 50.74 59.86 48.32 61.40 50.74 47.48 36.02 50.79 43.56 50.79 53.78 50.74 46.62 48.32 68.74 50.74
Wang_WST_task2_1 WangWST2026 148 52.847 60.54 79.38 53.16 45.90 51.68 59.96 50.95 58.60 49.63 69.14 65.21 46.31 37.64 53.16 44.90 51.68 44.66 50.95 40.82 49.63 81.64 65.21
Wang_WST_task2_2 WangWST2026 173 48.927 58.80 79.76 48.74 50.88 49.47 58.18 52.11 48.94 49.26 65.68 55.05 40.60 20.10 48.74 59.26 49.47 51.70 52.11 47.38 49.26 62.20 55.05
Wang_WST_task2_3 WangWST2026 145 53.137 60.89 79.98 53.05 45.84 50.84 61.20 53.16 58.54 49.68 69.58 64.05 46.68 35.64 53.05 46.72 50.84 47.62 53.16 41.24 49.68 80.66 64.05
Jiang_KY_task2_1 JiangKY2026 167 50.433 58.18 66.46 53.68 44.70 49.58 48.74 49.37 90.02 47.95 59.18 53.68 44.27 38.74 53.68 47.38 49.58 46.40 49.37 32.90 47.95 71.02 53.68
Jiang_KY_task2_2 JiangKY2026 160 51.165 60.01 67.76 54.84 51.04 49.58 50.56 49.11 90.08 47.84 55.28 50.00 45.38 41.52 54.84 49.66 49.58 49.88 49.11 32.10 47.84 67.78 50.00
Jiang_KY_task2_3 JiangKY2026 141 53.558 63.47 69.52 52.26 50.38 48.89 47.80 51.74 93.26 50.21 77.48 53.11 48.25 43.00 52.26 53.44 48.89 49.98 51.74 40.64 50.21 58.68 53.11
Jiang_KY_task2_4 JiangKY2026 144 53.174 65.66 74.60 51.79 54.38 49.42 48.94 50.84 93.24 50.26 75.78 52.11 46.45 38.24 51.79 53.44 49.42 45.64 50.84 41.06 50.26 60.54 52.11
Glitza_IKA_task2_1 GlitzaIKA2026 153 51.404 62.00 78.70 50.79 56.79 50.76 57.64 50.37 63.43 48.26 58.09 52.32 44.61 28.58 50.79 50.14 50.76 52.89 50.37 42.59 48.26 67.75 52.32
Glitza_IKA_task2_2 GlitzaIKA2026 147 52.856 63.34 80.80 52.53 56.77 50.74 63.74 55.95 61.76 49.42 58.60 54.89 45.54 30.24 52.53 51.90 50.74 50.68 55.95 42.02 49.42 71.86 54.89
Glitza_IKA_task2_3 GlitzaIKA2026 139 54.461 61.69 77.90 50.63 50.39 51.53 58.10 53.71 65.52 50.26 62.93 52.95 51.11 41.91 50.63 45.33 51.53 55.78 53.71 49.50 50.26 72.63 52.95
Glitza_IKA_task2_4 GlitzaIKA2026 164 50.773 55.72 78.92 51.63 46.64 49.84 51.74 48.47 56.06 50.68 54.16 48.37 47.52 35.40 51.63 48.41 49.84 54.88 48.47 44.56 50.68 63.86 48.37
Kajita_IND_task2_1 KajitaIND2026 159 51.170 54.12 65.68 53.63 55.68 49.84 51.44 51.11 51.00 49.84 49.62 50.32 48.76 47.96 53.63 42.94 49.84 54.60 51.11 49.06 49.84 50.75 50.32
Kajita_IND_task2_2 KajitaIND2026 161 51.130 54.12 65.24 53.68 54.53 49.76 51.66 51.32 51.26 50.21 50.36 50.32 48.55 49.66 53.68 41.70 49.76 53.76 51.32 49.15 50.21 50.21 50.32
Kajita_IND_task2_3 KajitaIND2026 151 51.780 52.51 60.14 53.58 47.17 49.76 51.32 51.53 50.44 50.11 55.32 50.26 51.84 61.14 53.58 50.82 49.76 55.22 51.53 54.42 50.11 41.78 50.26
Kajita_IND_task2_4 KajitaIND2026 158 51.188 54.15 65.66 53.79 55.42 49.95 51.58 51.05 51.12 49.89 49.70 50.32 48.74 48.38 53.79 42.46 49.95 54.46 51.05 49.34 49.89 50.69 50.32
Zhang_JAIST_task2_1 ZhangJAIST2026 47 61.026 69.81 80.00 54.42 68.18 51.00 68.20 64.37 63.38 53.47 71.34 69.00 57.17 44.90 54.42 53.82 51.00 66.38 64.37 52.72 53.47 79.52 69.00
Wang_UniS_task2_1 WangUniS2026 72 59.593 66.14 72.68 56.68 53.42 49.05 62.22 57.63 69.22 52.16 79.40 78.00 56.34 50.48 56.68 52.68 49.05 55.68 57.63 46.88 52.16 93.82 78.00
Wang_UniS_task2_2 WangUniS2026 143 53.384 68.53 79.78 48.53 56.36 49.26 71.06 65.53 64.20 51.95 76.74 65.11 42.65 23.36 48.53 33.88 49.26 63.30 65.53 52.74 51.95 98.40 65.11
Wang_UniS_task2_3 WangUniS2026 87 58.405 68.66 77.77 54.42 55.24 49.11 66.96 62.05 67.85 50.63 82.07 77.08 51.85 37.88 54.42 44.36 49.11 58.76 62.05 49.21 50.63 98.53 77.08
Wang_UniS_task2_4 WangUniS2026 92 57.968 68.72 78.13 54.21 55.32 48.97 67.27 62.24 67.41 50.61 82.05 75.95 50.95 36.06 54.21 43.15 48.97 59.11 62.24 49.52 50.61 98.82 75.95
Yang_XJU_task2_1 YangXJU2026 99 57.687 73.07 79.56 55.47 68.98 48.42 70.08 61.11 75.42 52.84 72.30 64.89 48.91 36.56 55.47 34.54 48.42 61.92 61.11 54.30 52.84 88.10 64.89
Yang_XJU_task2_2 YangXJU2026 85 58.624 59.66 48.24 50.79 55.36 49.21 60.20 61.58 64.06 51.11 78.18 61.05 62.61 60.34 50.79 53.70 49.21 63.18 61.58 57.02 51.11 88.46 61.05
Yang_XJU_task2_3 YangXJU2026 105 57.472 73.60 79.58 53.95 69.06 48.68 69.36 61.74 79.26 52.21 72.20 67.53 48.16 35.18 53.95 33.14 48.68 63.02 61.74 55.72 52.21 87.60 67.53
Yang_XJU_task2_4 YangXJU2026 62 60.152 72.15 73.42 52.21 66.68 48.21 65.48 62.95 76.08 52.68 81.50 73.37 54.54 43.98 52.21 42.24 48.21 61.98 62.95 54.90 52.68 91.66 73.37
SNU_task2_1 SNUtask22026 169 50.023 56.78 74.40 51.79 53.10 48.95 45.26 49.79 58.82 53.42 59.90 56.53 43.27 28.18 51.79 43.94 48.95 44.62 49.79 53.38 53.42 61.82 56.53
Huang_WHU_task2_1 HuangWHU2026 4 65.453 65.44 68.10 58.89 60.02 55.16 61.70 65.42 65.26 55.63 73.92 71.16 71.08 70.52 58.89 69.06 55.16 70.02 65.42 62.48 55.63 87.74 71.16
Huang_WHU_task2_2 HuangWHU2026 7 64.965 64.69 67.70 57.74 60.98 54.84 60.46 65.58 64.26 55.16 71.30 68.47 71.37 70.42 57.74 68.64 54.84 70.84 65.58 62.02 55.16 90.48 68.47
Huang_WHU_task2_3 HuangWHU2026 31 61.851 60.61 65.72 59.18 42.75 51.42 63.13 64.95 63.86 53.50 80.70 77.32 65.23 73.57 59.18 46.89 51.42 67.68 64.95 61.65 53.50 93.12 77.32
Huang_WHU_task2_4 HuangWHU2026 83 58.773 57.20 39.48 55.37 66.00 52.00 66.90 61.68 65.25 50.16 60.05 59.63 64.43 81.86 55.37 58.53 52.00 62.02 61.68 55.71 50.16 70.28 59.63
Morita_KM_task2_1 MoritaKM2026 117 56.956 70.24 81.12 53.11 57.68 53.53 70.78 52.79 79.72 52.79 67.34 61.00 49.81 35.06 53.11 49.48 53.53 50.70 52.79 48.00 52.79 90.14 61.00
Morita_KM_task2_2 MoritaKM2026 96 57.754 70.71 79.02 52.00 56.24 52.74 69.44 53.89 79.16 54.16 75.52 67.42 50.52 36.68 52.00 48.74 52.74 51.44 53.89 47.52 54.16 93.44 67.42
Morita_KM_task2_3 MoritaKM2026 97 57.722 65.41 69.20 51.86 57.93 52.58 65.29 55.13 64.05 53.56 72.47 65.61 53.72 38.65 51.86 52.49 52.58 59.06 55.13 53.37 53.56 80.14 65.61
Morita_KM_task2_4 MoritaKM2026 90 58.111 64.26 67.44 51.74 58.24 52.58 62.74 51.84 63.14 50.58 71.24 64.16 57.26 46.32 51.74 57.70 52.58 59.82 51.84 52.38 50.58 79.42 64.16
Wu_CUMT_task2_1 WuCUMT2026 13 63.292 63.98 61.44 56.84 49.38 52.37 71.30 60.05 63.64 53.37 84.18 76.68 67.81 57.72 56.84 65.60 52.37 70.50 60.05 60.36 53.37 96.02 76.68
Wu_CUMT_task2_2 WuCUMT2026 9 64.383 65.19 62.72 58.16 51.34 50.89 69.40 60.21 66.76 53.95 84.10 72.79 70.87 62.22 58.16 76.66 50.89 69.34 60.21 60.36 53.95 95.70 72.79
Wu_CUMT_task2_3 WuCUMT2026 5 65.181 67.66 66.24 58.58 59.86 55.42 69.90 61.37 62.66 53.32 84.54 70.58 69.54 61.74 58.58 70.86 55.42 71.74 61.37 57.62 53.32 97.14 70.58
Wu_CUMT_task2_4 WuCUMT2026 3 65.462 67.60 64.86 57.05 58.22 56.16 68.52 59.05 65.82 55.89 86.30 70.95 70.51 63.50 57.05 70.42 56.16 70.52 59.05 60.70 55.89 96.96 70.95
Yang_None_task2_1 YangNone2026 10 64.296 65.00 67.26 56.26 50.88 50.95 65.36 58.84 65.70 55.68 84.20 78.63 70.12 62.90 56.26 68.50 50.95 71.96 58.84 60.08 55.68 97.40 78.63
Yang_None_task2_2 YangNone2026 14 63.243 63.94 65.38 55.42 49.94 51.53 65.34 60.16 63.48 55.89 84.62 78.32 67.34 57.60 55.42 60.52 51.53 71.70 60.16 61.74 55.89 97.82 78.32
Yang_None_task2_3 YangNone2026 16 63.045 61.50 66.30 64.53 44.12 51.11 65.32 59.21 64.28 55.47 78.80 66.21 69.89 72.30 64.53 59.62 51.11 69.32 59.21 61.42 55.47 97.78 66.21
Yang_None_task2_4 YangNone2026 18 62.618 61.19 66.70 63.53 42.34 49.53 66.44 59.68 62.38 53.05 83.16 76.79 68.24 69.14 63.53 58.84 49.53 69.26 59.68 58.26 53.05 97.96 76.79
Zheng_HFUUAI_task2_1 ZhengHFUUAI2026 48 61.024 64.26 68.18 58.95 59.18 53.00 58.06 51.37 63.50 52.95 75.32 68.26 63.16 69.38 58.95 55.14 53.00 57.14 51.37 55.54 52.95 90.06 68.26
Zheng_HFUUAI_task2_2 ZhengHFUUAI2026 40 61.229 65.49 72.22 56.47 60.86 53.84 58.92 53.68 63.20 54.05 75.30 66.32 62.39 58.62 56.47 57.08 53.84 59.48 53.68 57.82 54.05 87.30 66.32
Zheng_HFUUAI_task2_3 ZhengHFUUAI2026 25 62.113 67.06 68.74 58.84 69.98 51.63 62.08 56.05 63.94 52.68 71.58 66.32 63.55 68.04 58.84 56.24 51.63 60.98 56.05 56.22 52.68 83.26 66.32
Zheng_HFUUAI_task2_4 ZhengHFUUAI2026 123 56.479 63.64 72.96 59.37 50.38 54.32 58.98 49.32 71.66 49.84 70.92 66.42 51.88 58.06 59.37 43.22 54.32 43.88 49.32 45.50 49.84 88.96 66.42
Zhou_HFUUDS_task2_1 ZhouHFUUDS2026 28 61.946 67.28 69.08 58.21 69.68 51.74 62.62 56.47 64.66 52.89 71.18 66.84 62.69 68.66 58.21 54.24 51.74 60.54 56.47 54.66 52.89 83.76 66.84
Zhou_HFUUDS_task2_2 ZhouHFUUDS2026 22 62.329 67.28 70.12 57.26 66.12 52.11 63.80 56.89 65.30 53.79 71.74 68.00 63.44 65.78 57.26 54.66 52.11 60.32 56.89 57.20 53.79 88.82 68.00
Zhou_HFUUDS_task2_3 ZhouHFUUDS2026 26 62.099 66.86 68.42 58.84 69.84 51.79 62.00 56.42 63.98 52.79 70.96 66.58 63.47 68.58 58.84 55.62 51.79 60.82 56.42 56.26 52.79 83.36 66.58
Zhou_HFUUDS_task2_4 ZhouHFUUDS2026 20 62.401 67.43 70.10 57.37 66.18 52.05 63.76 56.74 65.24 53.74 72.64 68.26 63.53 65.32 57.37 55.40 52.05 60.24 56.74 57.14 53.74 88.98 68.26
Zhu_FDA_task2_1 ZhuFDA2026 95 57.782 67.55 73.18 50.37 55.96 51.68 67.00 59.16 73.80 53.89 71.36 72.37 51.50 39.62 50.37 46.68 51.68 56.52 59.16 49.56 53.89 79.60 72.37
Zhu_FDA_task2_2 ZhuFDA2026 115 57.083 67.54 74.54 50.11 56.00 51.26 66.80 59.11 74.00 53.89 70.08 70.47 50.19 37.32 50.11 45.28 51.26 56.16 59.11 49.30 53.89 79.04 70.47
Zhu_FDA_task2_3 ZhuFDA2026 100 57.661 66.90 74.60 51.21 54.66 51.84 67.40 59.11 70.18 52.37 71.66 66.37 52.35 41.54 51.21 49.18 51.84 56.86 59.11 47.56 52.37 80.10 66.37
Zhu_FDA_task2_4 ZhuFDA2026 94 57.815 67.96 73.82 50.32 57.26 51.74 67.12 59.89 74.10 54.16 70.56 71.95 51.23 38.36 50.32 46.46 51.74 57.76 59.89 50.02 54.16 78.68 71.95
Huang_QWS_task2_1 HuangQWS2026 27 62.043 69.66 69.69 53.74 59.70 49.68 68.44 64.05 73.14 55.26 80.66 69.84 60.08 52.43 53.74 54.46 49.68 62.73 64.05 52.68 55.26 92.04 69.84
Wang_Liu_SuzhouDongyuan_task2_1 WangLiu2026 146 53.075 61.86 74.80 51.84 55.92 49.42 60.42 55.32 59.96 50.37 61.18 54.11 47.23 38.24 51.84 37.52 49.42 50.88 55.32 52.28 50.37 70.08 54.11
Wang_Liu_SuzhouDongyuan_task2_2 WangLiu2026 163 50.898 60.86 79.26 52.63 59.08 49.00 55.66 50.79 58.82 50.63 56.66 52.68 43.59 33.90 52.63 36.82 49.00 45.78 50.79 51.98 50.63 58.90 52.68
Wang_Liu_SuzhouDongyuan_task2_3 WangLiu2026 168 50.208 59.55 78.78 52.95 58.74 49.11 55.58 50.79 58.02 50.58 52.58 50.95 42.94 33.96 52.95 37.44 49.11 45.48 50.79 51.98 50.58 52.44 50.95
Wang_Liu_SuzhouDongyuan_task2_4 WangLiu2026 152 51.588 59.44 74.54 51.26 54.78 49.58 60.50 54.68 58.12 50.26 53.44 50.89 45.81 36.84 51.26 38.00 49.58 51.10 54.68 52.46 50.26 58.66 50.89
Qian_nivic_task2_1 Qiannivic2026 23 62.311 62.79 72.36 61.21 44.08 49.68 65.32 58.05 65.84 55.68 79.20 70.37 66.46 70.50 61.21 50.62 49.68 70.68 58.05 60.16 55.68 94.98 70.37
Qian_nivic_task2_2 Qiannivic2026 8 64.428 64.74 71.04 62.47 45.66 50.26 68.64 58.42 69.70 59.26 81.02 76.37 68.87 74.76 62.47 53.68 50.26 70.32 58.42 63.34 59.26 94.50 76.37
Qian_nivic_task2_3 Qiannivic2026 42 61.180 60.44 66.86 58.37 39.28 51.00 68.20 56.21 66.14 53.42 79.80 67.47 67.22 71.70 58.37 51.74 51.00 71.02 56.21 60.38 53.42 95.58 67.47
Qian_nivic_task2_4 Qiannivic2026 12 63.447 62.33 67.30 62.16 41.16 51.26 69.16 55.84 70.24 58.37 80.88 73.74 69.44 74.28 62.16 55.72 51.26 70.22 55.84 63.60 58.37 94.04 73.74
Jeong_Medisensing_task2_1 JeongMedisensing2026 119 56.854 67.59 78.50 52.58 53.76 49.79 65.76 57.68 69.58 51.26 76.56 74.95 49.72 35.06 52.58 47.68 49.79 56.12 57.68 45.94 51.26 87.08 74.95
Jeong_Medisensing_task2_2 JeongMedisensing2026 135 55.020 66.65 78.56 51.26 52.86 49.32 65.26 58.84 69.12 52.21 73.66 70.00 46.56 29.28 51.26 44.54 49.32 57.18 58.84 46.98 52.21 83.22 70.00
Jeong_Medisensing_task2_3 JeongMedisensing2026 125 56.139 66.70 78.72 52.53 53.62 48.95 64.50 57.95 67.84 51.79 74.84 69.89 49.07 34.60 52.53 43.94 48.95 57.62 57.95 47.46 51.79 84.70 69.89
Jeong_Medisensing_task2_4 JeongMedisensing2026 106 57.436 67.27 76.90 53.11 53.48 50.42 63.89 55.11 69.62 50.61 79.33 77.37 51.37 40.38 53.11 50.21 50.42 52.35 55.11 44.63 50.61 89.68 77.37
Zhou_XAUAT_task2_1 ZhouXAUAT2026 37 61.371 55.96 57.18 56.58 40.26 52.42 55.18 54.58 65.32 57.95 73.54 66.53 73.86 75.54 56.58 80.70 52.42 63.20 54.58 67.14 57.95 88.10 66.53
Kim_CAU_task2_1 KimCAU2026 101 57.581 63.60 75.90 60.58 49.42 47.79 62.44 60.32 62.66 52.58 75.60 63.47 53.70 48.54 60.58 42.40 47.79 56.60 60.32 49.14 52.58 91.72 63.47
Kim_CAU_task2_2 KimCAU2026 66 59.854 58.64 62.18 56.21 43.52 48.26 58.38 61.21 61.96 54.21 77.28 61.16 66.05 79.36 56.21 58.34 48.26 56.16 61.21 57.62 54.21 92.66 61.16
Kim_CAU_task2_3 KimCAU2026 86 58.484 63.11 75.06 60.21 51.10 47.68 60.82 61.21 62.92 52.63 71.42 61.74 56.73 55.76 60.21 44.16 47.68 60.14 61.21 50.70 52.63 89.22 61.74
Kim_CAU_task2_4 KimCAU2026 81 58.922 63.12 75.00 60.68 50.44 47.79 60.52 61.42 63.04 52.00 73.14 63.32 57.71 57.06 60.68 46.16 47.79 58.36 61.42 52.02 52.00 90.10 63.32
Zeng_BUCT_task2_1 ZengBUCT2026 165 50.716 48.16 64.24 52.00 38.42 48.16 52.72 51.58 44.18 52.16 48.50 51.26 53.26 39.20 52.00 57.08 48.16 56.98 51.58 53.72 52.16 68.06 51.26
Zeng_BUCT_task2_2 ZengBUCT2026 171 49.310 51.31 55.04 52.21 56.50 51.11 47.90 49.47 46.70 49.00 51.86 50.21 46.51 56.72 52.21 44.42 51.11 41.44 49.47 45.48 49.00 47.08 50.21
Zeng_BUCT_task2_3 ZengBUCT2026 172 49.267 48.79 54.78 50.37 55.00 53.16 37.78 49.53 48.02 47.68 53.32 50.89 48.78 51.64 50.37 49.16 53.16 47.82 49.53 42.30 47.68 54.82 50.89
Zeng_BUCT_task2_4 ZengBUCT2026 162 51.109 49.30 63.20 52.00 38.10 47.95 55.36 51.79 44.54 52.37 53.08 50.84 53.25 39.74 52.00 57.20 47.95 57.72 51.79 53.60 52.37 65.46 50.84
Yang_NJU_task2_1 YangNJU2026 43 61.174 59.88 65.76 57.11 42.10 48.16 56.36 51.16 69.82 60.84 80.20 76.74 67.19 77.32 57.11 49.92 48.16 58.18 51.16 75.66 60.84 90.52 76.74
Yang_NJU_task2_2 YangNJU2026 57 60.350 59.83 62.64 54.37 42.60 48.16 56.92 51.16 71.76 61.16 79.16 73.68 65.49 72.64 54.37 47.86 48.16 56.86 51.16 76.16 61.16 91.18 73.68
Kim_KATECH_task2_1 KimKATECH2026 73 59.498 67.22 76.10 58.68 56.06 57.58 66.12 55.00 71.50 50.58 69.96 66.47 55.36 51.28 58.68 50.18 57.58 55.00 55.00 46.34 50.58 89.84 66.47
Kim_KATECH_task2_2 KimKATECH2026 80 58.985 66.78 76.56 58.68 54.42 57.21 65.56 55.26 72.56 49.84 69.42 65.58 54.65 51.48 58.68 48.88 57.21 54.58 55.26 44.88 49.84 90.82 65.58
Kim_KATECH_task2_3 KimKATECH2026 61 60.210 70.54 77.36 59.58 57.00 56.74 71.92 54.42 76.74 50.58 74.18 67.26 55.05 50.24 59.58 50.30 56.74 53.62 54.42 46.40 50.58 92.32 67.26
Kim_KATECH_task2_4 KimKATECH2026 77 59.388 67.26 76.42 59.00 56.00 57.42 65.86 55.37 71.80 50.00 70.00 66.63 55.06 51.20 59.00 49.82 57.42 54.64 55.37 45.96 50.00 89.76 66.63
Guan_GISP@HEU_task2_1 GuanGISP@HEU2026 79 59.008 67.79 77.90 61.11 59.92 53.53 67.90 59.11 63.98 48.74 72.06 68.53 53.55 49.58 61.11 52.52 53.53 57.86 59.11 39.98 48.74 84.32 68.53
Krag_AAU_task2_1 KragAAU2026 38 61.328 73.10 75.66 58.47 70.96 50.53 69.28 55.37 70.54 53.21 80.14 73.84 56.24 50.04 58.47 48.88 50.53 52.22 55.37 53.42 53.21 94.46 73.84
Krag_AAU_task2_2 KragAAU2026 104 57.472 64.88 73.21 53.76 64.16 50.55 60.26 56.71 62.73 52.08 65.42 66.16 53.42 45.20 53.76 51.88 50.55 59.34 56.71 50.45 52.08 64.42 66.16
Krag_AAU_task2_3 KragAAU2026 116 56.961 65.15 71.36 54.74 67.22 49.21 60.18 55.95 63.34 52.58 64.74 64.32 52.32 51.42 54.74 42.34 49.21 57.90 55.95 50.96 52.58 64.08 64.32
Krag_AAU_task2_4 KragAAU2026 19 62.588 71.39 73.23 58.24 68.58 54.61 66.55 54.13 71.38 54.50 78.34 74.58 59.59 57.53 58.24 57.16 54.61 52.60 54.13 54.43 54.50 85.84 74.58
Moradi_JKU_task2_1 MoradiJKU2026 52 60.795 68.46 77.02 53.21 57.04 53.95 71.08 65.53 73.86 53.53 67.08 61.16 58.09 46.04 53.21 48.58 53.95 70.56 65.53 59.78 53.53 77.70 61.16
Moradi_JKU_task2_2 MoradiJKU2026 60 60.214 68.26 78.22 53.00 55.92 54.63 71.68 65.21 73.72 53.68 66.38 61.84 56.46 43.38 53.00 47.62 54.63 69.80 65.21 58.96 53.68 75.64 61.84



System characteristics

Summary of the submitted system characteristics.

Rank Submission
Code 		Technical
Report 		Classifier System
Complexity 		Acoustic
Feature 		Data
Augmentation 		Decision Making System
Embeddings 		Subsystem
Conut 		External
Data Usage 		Front End
System
137 DCASE2026_baseline_task2_MAHALA DCASE2026baseline2026 AE log-mel energies
68 DCASE2026_baseline_task2_MSE DCASE2026baseline2026 AE log-mel energies
121 Ozeki_MELCO_task2_1 OzekiMELCO2026 kNN Mel spectrogram CED pre-trained model, embeddings Normalization, Patchification, Linear filter
107 Ozeki_MELCO_task2_2 OzekiMELCO2026 kNN log-mel filterbank BEATs average Normalization, log-Mel filterbank extraction
59 Ozeki_MELCO_task2_3 OzekiMELCO2026 AE Mel spectrogram Linear filter
49 Ozeki_MELCO_task2_4 OzekiMELCO2026 AE log-mel filterbank domain BEATs Normalization, log-Mel filterbank extraction
89 Qian_SJTU_task2_1 QianSJTU2026 pre-trained models log-mel filterbank SpecAugment BEATs
54 Qian_SJTU_task2_2 QianSJTU2026 pre-trained models log-mel filterbank SpecAugment FISHER-small
91 Qian_SJTU_task2_3 QianSJTU2026 ensemble, pre-trained models log-mel filterbank SpecAugment Bayesian optimization 3 BEATs
88 Qian_SJTU_task2_4 QianSJTU2026 ensemble, pre-trained models log-mel filterbank SpecAugment Bayesian optimization 7 BEATs
55 Qian_VUILabs_task2_1 QianVUILabs2026 pre-trained models log-mel filterbank SpecAugment BEATs
78 Qian_VUILabs_task2_2 QianVUILabs2026 pre-trained models log-mel filterbank SpecAugment BEATs
56 Qian_VUILabs_task2_3 QianVUILabs2026 ensemble, pre-trained models log-mel filterbank SpecAugment Bayesian optimization 5 BEATs
63 Qian_VUILabs_task2_4 QianVUILabs2026 ensemble, pre-trained models log-mel filterbank SpecAugment Bayesian optimization 2 BEATs
133 Zhang_XJTLU_task2_1 ZhangXJTLU2026 reliability gating, score ensemble, train-normal density scoring frozen audio embeddings, train-normal kNN distance scores fixed train-normal score threshold CED-small, Dasheng, EAT 3 pre-trained embeddings near channel, mono mean channel
134 Zhang_XJTLU_task2_2 ZhangXJTLU2026 reliability gating, score ensemble, train-normal density scoring frozen audio embeddings, train-normal kNN distance scores fixed train-normal score threshold CED-small, Dasheng, EAT 3 pre-trained embeddings near channel, mono mean channel
132 Zhang_XJTLU_task2_3 ZhangXJTLU2026 reliability gating, score ensemble, train-normal density scoring frozen audio embeddings, train-normal kNN distance scores fixed train-normal score threshold CED-small, Dasheng, EAT 3 pre-trained embeddings near channel, mono mean channel
131 Zhang_XJTLU_task2_4 ZhangXJTLU2026 reliability gating, score ensemble, train-normal density scoring frozen audio embeddings, train-normal kNN distance scores fixed train-normal score threshold CED-small, Dasheng, EAT 3 pre-trained embeddings near channel, mono mean channel
58 Chang_Surrey_task2_1 ChangSurrey2026 ensemble, k-NN (cosine), Mahalanobis distance condition-monitoring indicators, inter-channel spatial features, log-mel energies, raw waveform weighted maximum M2D, M2D-CLAP, BEATs, EAT, domain-adapted M2D (further pre-trained on machine audio) 10 pre-trained model 2-mic Wiener spatial noise cancellation, band-pass filtering (4-7.9kHz)
76 Chang_Surrey_task2_2 ChangSurrey2026 ensemble, k-NN (cosine), Mahalanobis distance condition-monitoring indicators, inter-channel spatial features, log-mel energies, raw waveform weighted average M2D, M2D-CLAP, BEATs, EAT, domain-adapted M2D (further pre-trained on machine audio) 10 pre-trained model 2-mic Wiener spatial noise cancellation, band-pass filtering (4-7.9kHz)
112 Chang_Surrey_task2_3 ChangSurrey2026 ensemble, k-NN (cosine), Mahalanobis distance condition-monitoring indicators, inter-channel spatial features, log-mel energies, raw waveform weighted average BEATs, EAT 7 pre-trained model 2-mic Wiener spatial noise cancellation, band-pass filtering (4-7.9kHz)
142 Chang_Surrey_task2_4 ChangSurrey2026 ensemble, Mahalanobis distance condition-monitoring indicators, inter-channel spatial features, log-mel energies, raw waveform weighted average 2
45 Zhang_SATLab_task2_1 ZhangSATLab2026 diffusion, pre-trained models log-mel filterbank SpecAugment BEATs BEATs (pre-trained model)
71 Zhang_SATLab_task2_2 ZhangSATLab2026 diffusion, ensemble, pre-trained models log-mel filterbank SpecAugment weighted average (Bayesian optimization) BEATs 7 BEATs (pre-trained model)
39 Zhang_SATLab_task2_3 ZhangSATLab2026 diffusion, ensemble, pre-trained models log-mel filterbank SpecAugment weighted average (Bayesian optimization) BEATs 9 BEATs (pre-trained model)
44 Zhang_SATLab_task2_4 ZhangSATLab2026 diffusion, ensemble, pre-trained models log-mel filterbank SpecAugment weighted average (Bayesian optimization) BEATs 12 BEATs (pre-trained model)
35 Fan_WISTLAB_task2_1 FanWISTLAB2026 pre-trained models log-mel filterbank, STFT SpecAugment BEATs
33 Fan_WISTLAB_task2_2 FanWISTLAB2026 ensemble, pre-trained models log-mel filterbank, STFT SpecAugment Bayesian optimization 6 BEATs
34 Fan_WISTLAB_task2_3 FanWISTLAB2026 ensemble, pre-trained models log-mel filterbank, STFT SpecAugment Bayesian optimization 7 BEATs
17 Fan_WISTLAB_task2_4 FanWISTLAB2026 ensemble, pre-trained models log-mel filterbank, STFT SpecAugment Bayesian optimization 2 BEATs
98 Jiang_AITHU_task2_1 JiangAITHU2026 Mahalanobis distance, pre-trained models log-mel filterbank, STFT SpecAugment BEATs
82 Jiang_AITHU_task2_2 JiangAITHU2026 ensemble, pre-trained models log-mel filterbank, STFT SpecAugment Bayesian optimization 10 BEATs
21 Jiang_AITHU_task2_3 JiangAITHU2026 ensemble, pre-trained models log-mel filterbank, STFT SpecAugment Bayesian optimization 3 BEATs
41 Jiang_AITHU_task2_4 JiangAITHU2026 ensemble, Mahalanobis distance, pre-trained models log-mel filterbank, STFT SpecAugment Bayesian optimization 7 BEATs
30 Zhang_THUEE_task2_1 ZhangTHUEE2026 kNN, pre-trained models log-mel filterbank, STFT SpecAugment BEATs
51 Zhang_THUEE_task2_2 ZhangTHUEE2026 ensemble, kNN, Mahalanobis distance, pre-trained models log-mel filterbank, STFT SpecAugment Bayesian optimization 16 BEATs
24 Zhang_THUEE_task2_3 ZhangTHUEE2026 ensemble, kNN, Mahalanobis distance, pre-trained models log-mel filterbank, STFT SpecAugment Bayesian optimization 23 BEATs
11 Zhang_THUEE_task2_4 ZhangTHUEE2026 ensemble, kNN, pre-trained models log-mel filterbank, STFT SpecAugment Bayesian optimization 7 BEATs
111 Huang_CQUPT_task2_1 HuangCQUPT2026 frozen pre-trained audio encoders, kNN, score fusion raw waveform train-normal z-score alignment, weighted average Dasheng, EAT-AS2M, EAT-large, OpenL3, SSLAM 5 pre-trained audio embeddings STFT soft mask
124 Huang_CQUPT_task2_2 HuangCQUPT2026 frozen pre-trained audio encoder, kNN raw waveform EAT-large pre-trained audio embeddings
122 Huang_CQUPT_task2_3 HuangCQUPT2026 frozen pre-trained audio encoders, kNN, score fusion raw waveform train-normal z-score alignment, weighted average Dasheng, EAT-large, OpenL3 3 pre-trained audio embeddings STFT soft mask
109 Huang_CQUPT_task2_4 HuangCQUPT2026 frozen pre-trained audio encoders, kNN, score fusion raw waveform train-normal z-score alignment, weighted average Dasheng, EAT-large, SSLAM 3 pre-trained audio embeddings STFT soft mask
75 Xie_SHU_task2_1 XieSHU2026 kNN log-mel energies minimum EAT
65 Xie_SHU_task2_2 XieSHU2026 kNN log-mel energies minimum EAT
84 Xie_SHU_task2_3 XieSHU2026 kNN log-mel energies minimum EAT
93 Xie_SHU_task2_4 XieSHU2026 kNN log-mel energies minimum EAT
127 Moon_Independent_task2_1 MoonIndependent2026 AE, KMeans, LOF, Mahalanobis distance, MLP, PatchCore, weighted ensemble energy ratio, GCC-PHAT, inter-channel coherence, log-mel energies, mel-band cross-spectrum, MFCC, spectral kurtosis none thresholding of the final anomaly score using a fixed threshold determined before evaluation BEATs, LAION-CLAP 5 pretrained BEATs and LAION-CLAP models listed in the official allowed external resources; no external training dataset directly used Wiener filtering, spectral subtraction, PCA, median-MAD score normalization
32 Xia_NEU_task2_1 XiaNEU2026 cosine k-nearest-neighbor anomaly scoring FFT magnitude spectrum and STFT magnitude spectra mixup fixed 0.9 normal-train-score quantile threshold 3 Raw wav ch1, Wiener ch2
69 Xia_NEU_task2_2 XiaNEU2026 cosine k-nearest-neighbor anomaly scoring pre-trained audio embeddings mixup a fixed 0.9 normal-train-score quantile threshold BEATs BEATs pre-trained audio representation model Raw wav ch1
67 Xia_NEU_task2_3 XiaNEU2026 cosine k-nearest-neighbor anomaly scoring pre-trained audio embeddings mixup a fixed 0.9 normal-train-score quantile threshold BEATs BEATs pre-trained audio representation model Raw wav ch1
46 Xia_NEU_task2_4 XiaNEU2026 BEATs embedding extractor with rank-128 LoRA modules, sub-cluster AdaCos classifier, and cosine k-nearest-neighbor anomaly scoring pre-trained audio embeddings mixup continuous cosine nearest-neighbor anomaly score (0.9 normal-train quantile threshold) BEATs BEATs pre-trained audio representation model Raw wav ch1
50 XingWu_MCPX_task2_1 XingWuMCPX2026 EAT/FISHER embeddings with cosine KNN audio embeddings mixup normal-training score threshold with an expected 5% abnormal rate on normal training clips EAT and FISHER models 3 EAT and FISHER pretrained models Ch1 waveform; optional Ch2-guided preprocessing
36 XingWu_MCPX_task2_2 XingWuMCPX2026 EAT/FISHER embeddings with cosine KNN audio embeddings mixup normal-training score threshold with an expected 5% abnormal rate on normal training clips EAT and FISHER models 6 EAT and FISHER pretrained models Ch1 waveform; optional Ch2-guided preprocessing
74 XingWu_MCPX_task2_3 XingWuMCPX2026 EAT ArcFace embedding with cosine KNN audio embeddings mixup normal-training score threshold with an expected 5% abnormal rate on normal training clips EAT EAT pretrained model Ch1 waveform; optional Ch2-guided preprocessing
29 XingWu_MCPX_task2_4 XingWuMCPX2026 EAT/FISHER embeddings with cosine KNN audio embeddings mixup normal-training score threshold with an expected 5% abnormal rate on normal training clips EAT and FISHER models 9 EAT and FISHER pretrained models Ch1 waveform; optional Ch2-guided preprocessing
175 Zhou_SUMERUZOO_task2_1 ZhouSUMERUZOO2026 cosine distance, Mahalanobis distance log-mel energies none maximum BEATs 4 pre-trained model none
174 Zhou_SUMERUZOO_task2_3 ZhouSUMERUZOO2026 Mahalanobis distance log-mel energies, spectral and temporal statistics none none
110 Kwon_KIST_task2_1 KwonKIST2026 ArcFace, ensemble, kNN, LoRA fine-tuning, Mahalanobis distance log-mel energies mixup weighted average BEATs, fine-tuned BEATs (LoRA) 3 pre-trained model dual-microphone magnitude spectral subtraction
114 Kwon_KIST_task2_2 KwonKIST2026 ensemble, Mahalanobis distance log-mel energies weighted average BEATs 2 pre-trained model dual-microphone magnitude spectral subtraction
120 Kwon_KIST_task2_3 KwonKIST2026 ensemble, Mahalanobis distance log-mel energies weighted average BEATs 2 pre-trained model
113 Kwon_KIST_task2_4 KwonKIST2026 ensemble, Mahalanobis distance log-mel energies weighted average BEATs, Dasheng 2 pre-trained model
2 Fujimura_MERL_task2_1 FujimuraMERL2026 log-mel energies mixup average BEATs, EAT 6 pre-training
15 Fujimura_MERL_task2_2 FujimuraMERL2026 log-mel energies average BEATs, EAT, EAT large 9 pre-training
1 Fujimura_MERL_task2_3 FujimuraMERL2026 log-mel energies mixup average BEATs 3 pre-training
6 Fujimura_MERL_task2_4 FujimuraMERL2026 log-mel energies mixup average EAT 6 pre-training
118 Noh_CBNU_task2_1 NohCBNU2026 AE, ensemble, kNN, Mahalanobis distance, masked spectrogram modeling, PaDiM, sub-cluster AdaCos cross-channel coherence, log-mel energies, modulation spectrum weighted average BEATs 8 pre-trained model
130 Jeong_KETI_task2_1 JeongKETI2026 masked spectrogram autoencoder (CNN), kNN density far-context ridge-residual descriptor, log-mel energies, temporal modulation descriptor spectrogram patch masking
140 Zarrouky_IR_task2_1 ZarroukyIR2026 Mahalanobis distance with Ledoit-Wolf shrinkage CQT, log-mel energies median threshold 2 dual-channel (ch1, ch2, ch1-ch2)
136 Lei_CRRC_task2_1 LeiCRRC2026 AE, selective Mahalanobis distance log-mel energies near microphone channel
128 Lei_CRRC_task2_2 LeiCRRC2026 AE, ensemble, kNN, selective Mahalanobis distance log-mel energies, M2D-AS embeddings weighted average M2D-AS 2 pre-trained model, embeddings near microphone channel
149 Lei_CRRC_task2_3 LeiCRRC2026 AE, selective Mahalanobis distance log-mel energies least-squares far-channel noise residual
108 Lei_CRRC_task2_4 LeiCRRC2026 AE, ArcFace, attention LoRA, audio transformers, ensemble, local-density kNN, PCA whitening log-mel energies and LoRA-adapted transformer hidden states weighted anomaly-score fusion (per-machine normal-reference quantile threshold) CED-Base, DualMic-EAT, M2D-AS 4 pre-trained embedding models and parameter-efficient fine-tuning synchronized near and far microphone channels
126 Tsz_HFU_task2_1 TszHFU2026 DCASE2023T2 autoencoder, per-machine training on normal clips log-mel energies gamma-fit threshold on train-normal scores (q=0.9) log-mel spectrogram (128 mels, 5-frame stack)
156 Tsz_HFU_task2_2 TszHFU2026 ArcFace multi-class training on composite labels; rank-normalized KNN+GMM fusion scoring BEATs encoder patch sequence, generalized mean pooling gaussian noise, time shift gamma-fit threshold on train-normal fusion scores (q=0.9) 768-d clip embedding after top-4-layer BEATs fine-tune 2 BEATs iter3+AS2M (AudioSet) pretrained checkpoint frozen/finetuned BEATs (near-mic mono, 10s clip)
103 Tsz_HFU_task2_3 TszHFU2026 ASP + ArcFace multi-task (machine + attribute heads) BEATs encoder patch sequence gamma-fit threshold on train-normal scores (q=0.9) Attentive Statistics Pooling + BN BEATs iter3+AS2M finetuned checkpoint (AudioSet) frozen BEATs (near-mic mono waveform)
170 Tsz_HFU_task2_4 TszHFU2026 CNN+MSFE, discriminative embedding, FiLM+Conformer, prototype min-distance scoring, SGFF fusion log-mel energies, spatial cues mixup, target oversampling gamma-fit threshold on train-normal proto scores (q=0.9) BEATs (frozen, near-mic waveform) pre-trained BEATs checkpoint (AudioSet) stereo log-mel + spatial cues (ILD/ITD/IPD/MS), PCEN optional path
70 Kim_LUDO_task2_1 KimLUDO2026 kNN, pre-trained models log-mel energies MemMixup average SSLAM pre-trained model
64 Kim_LUDO_task2_2 KimLUDO2026 kNN, pre-trained models log-mel energies MemMixup average BEATs iter3, DaSheng-base, SSLAM 3 pre-trained model
53 Kim_LUDO_task2_3 KimLUDO2026 kNN, pre-trained models log-mel energies MemMixup SSLAM pre-trained model
102 Kim_LUDO_task2_4 KimLUDO2026 kNN, pre-trained models log-mel energies MemMixup average AudioMAE++ tiny pre-trained model
155 Balozi_RISE_task2_1 BaloziRISE2026 ensemble, kNN raw waveform MemMixup weighted average AST, BEATs (LoRA r=16), BEATs (LoRA r=32), CLAP, CLAP (LoRA), WavLM 6 pre-trained model
154 Balozi_RISE_task2_2 BaloziRISE2026 ensemble, kNN raw waveform MemMixup weighted average CLAP, WavLM, BEATs (LoRA r=16, all12), BEATs (LoRA r=32, all12), CLAP (LoRA, all12), AST (all12) 6 pre-trained model
166 Balozi_RISE_task2_3 BaloziRISE2026 ensemble, kNN raw waveform MemMixup weighted average AST (all12), BEATs (LoRA r=16), BEATs (LoRA r=32), CLAP, CLAP (LoRA, all12), WavLM 6 pre-trained model
157 Balozi_RISE_task2_4 BaloziRISE2026 ensemble, kNN raw waveform MemMixup weighted average BEATs (LoRA r=16), BEATs (LoRA r=32), CLAP, CLAP (LoRA), WavLM 5 pre-trained model
129 Mei_FDID_task2_1 MeiFDID2026 GMM, UBM likelihood ratio handcrafted features, log-mel energies, nonlinear features likelihood ratio 2
138 Mei_FDID_task2_2 MeiFDID2026 GMM, UBM likelihood ratio handcrafted features, log-mel energies, nonlinear features likelihood ratio 2
150 Mei_FDID_task2_3 MeiFDID2026 GMM handcrafted features, log-mel energies, nonlinear features negative log-likelihood
148 Wang_WST_task2_1 WangWST2026 AE, baseline MAHALA, baseline MSE, EAT, ensemble, GMM, LOF, prototype log-mel energies percentile threshold EAT pre-trained EAT model
173 Wang_WST_task2_2 WangWST2026 AE, EAT, ensemble, LOF, prototype log-mel energies percentile threshold EAT pre-trained EAT model
145 Wang_WST_task2_3 WangWST2026 AE, baseline MAHALA, baseline MSE, EAT, ensemble, GMM, LOF, prototype, stereo AE log-mel energies percentile threshold EAT pre-trained EAT model
167 Jiang_KY_task2_1 JiangKY2026 CNN, cross-attention log-mel energies SpecAugment average 20
160 Jiang_KY_task2_2 JiangKY2026 CNN, cross-attention log-mel energies SpecAugment average 10
141 Jiang_KY_task2_3 JiangKY2026 CNN, cross-attention log-mel energies SpecAugment average 20
144 Jiang_KY_task2_4 JiangKY2026 CNN, cross-attention log-mel energies SpecAugment average 10
153 Glitza_IKA_task2_1 GlitzaIKA2026 kNN spectrogram threshold FISHER
147 Glitza_IKA_task2_2 GlitzaIKA2026 GMM spectrogram threshold FISHER
139 Glitza_IKA_task2_3 GlitzaIKA2026 kNN log-mel energies threshold
164 Glitza_IKA_task2_4 GlitzaIKA2026 GMM log-mel energies threshold
159 Kajita_IND_task2_1 KajitaIND2026 kNN, Mahalanobis distance, PCA residual scoring, rank-based ensemble, robust statistics clip-level acoustic statistics, spectral descriptors, periodicity descriptors, two-channel near/far relation descriptors train-normal 99th percentile threshold 54 robust clip-level normalization
161 Kajita_IND_task2_2 KajitaIND2026 kNN, Mahalanobis distance, PCA residual scoring, rank-based ensemble, robust statistics clip-level acoustic statistics, spectral descriptors, periodicity descriptors, two-channel near/far relation descriptors train-normal 99th percentile threshold 54 robust clip-level normalization
151 Kajita_IND_task2_3 KajitaIND2026 kNN, Mahalanobis distance, PCA residual scoring, rank-based ensemble, robust statistics clip-level acoustic statistics, spectral descriptors, periodicity descriptors, two-channel near/far relation descriptors train-normal 99th percentile threshold 54 robust clip-level normalization
158 Kajita_IND_task2_4 KajitaIND2026 kNN, Mahalanobis distance, PCA residual scoring, rank-based ensemble, robust statistics clip-level acoustic statistics, spectral descriptors, periodicity descriptors, two-channel near/far relation descriptors train-normal 99th percentile threshold 54 robust clip-level normalization
47 Zhang_JAIST_task2_1 ZhangJAIST2026 kNN raw waveform weighted average BEATs, CED 4 pre-trained model
72 Wang_UniS_task2_1 WangUniS2026 kNN log-mel energies mixup EAT-base
143 Wang_UniS_task2_2 WangUniS2026 kNN log-mel energies mixup BEATs
87 Wang_UniS_task2_3 WangUniS2026 KNN log-mel energies mixup BEATs，EAT 2
92 Wang_UniS_task2_4 WangUniS2026 kNN log-mel energies mixup BEATs, EAT 2
99 Yang_XJU_task2_1 YangXJU2026 EAT log-mel energies mixup, SpecAugment
85 Yang_XJU_task2_2 YangXJU2026 BEATs log-mel energies mixup, SpecAugment
105 Yang_XJU_task2_3 YangXJU2026 EAT log-mel energies mixup, SpecAugment ANC
62 Yang_XJU_task2_4 YangXJU2026 BEATs, EAT log-mel energies mixup, SpecAugment average 10
169 SNU_task2_1 SNUtask22026 contrastive learning, ResNet34, attention pooling, temporal convolutional network, domain adversarial neural network (DANN) RGB log-Mel spectrogram random resized crop, random horizontal flip, color jitter, random grayscale ResNet34 ImageNet pretrained embedding, projection head embedding, fused spatial-temporal embedding 1 log-Mel spectrogram, RGB colormap conversion, image resizing
4 Huang_WHU_task2_1 HuangWHU2026 ensemble, kNN, LoRA, Mahalanobis distance log-mel energies average ATST-Frame, BEATs, EAT-large, SSLAM 9 pre-trained model, embeddings
7 Huang_WHU_task2_2 HuangWHU2026 ensemble, kNN, LoRA, Mahalanobis distance log-mel energies waveform perturbation average ATST-Frame, BEATs, EAT-large, SSLAM 15 pre-trained model, embeddings
31 Huang_WHU_task2_3 HuangWHU2026 cosine similarity, ensemble, GMM, kNN, local density normalization, Mahalanobis distance log-mel energies mixup rank-normalized weighted average BEATs, EAT-large, SSLAM 8 pre-trained model, LoRA fine-tuning
83 Huang_WHU_task2_4 HuangWHU2026 ensemble, kNN, local density normalization SSL transformer layer embeddings rank_mean (branch fusion), gamma-fit threshold on train normal scores (90th percentile) BEATs_ft1, CED-tiny, EAT-large, M2D-CLAP, SSLAM 5 pre-trained model inter-channel SNR profiling, stereo channel alignment, SNR-aware enhancement, layer EMA smoothing
117 Morita_KM_task2_1 MoritaKM2026 log-mel energies Temporal crop augmentation EAT Pre-trained model Coherence and power-ratio mask
96 Morita_KM_task2_2 MoritaKM2026 log-mel energies Temporal crop augmentation average EAT 4 Pre-trained model Coherence and power-ratio mask
97 Morita_KM_task2_3 MoritaKM2026 kNN, ZCA whitening band energy ratio, impulsiveness, log-mel energies, MFCC, spectral contrast, stereo coherence profile mixup weighted average EAT 2 Pre-trained model
90 Morita_KM_task2_4 MoritaKM2026 kNN, ZCA whitening band energy ratio, impulsiveness, MFCC, spectral contrast, stereo coherence profile mixup
13 Wu_CUMT_task2_1 WuCUMT2026 kNN log-mel energies mixup Kalman filter
9 Wu_CUMT_task2_2 WuCUMT2026 kNN log-mel energies mixup Kalman filter
5 Wu_CUMT_task2_3 WuCUMT2026 kNN log-mel energies mixup Kalman filter
3 Wu_CUMT_task2_4 WuCUMT2026 kNN log-mel energies mixup Kalman filter
10 Yang_None_task2_1 YangNone2026 kNN log-mel energies mixup
14 Yang_None_task2_2 YangNone2026 kNN log-mel energies mixup
16 Yang_None_task2_3 YangNone2026 kNN log-mel energies mixup
18 Yang_None_task2_4 YangNone2026 kNN log-mel energies mixup
48 Zheng_HFUUAI_task2_1 ZhengHFUUAI2026 kNN, score fusion BEATs LoRA embeddings, spatial statistics noise mixing, random gain, time shift weighted average of normalized anomaly scores BEATs 3 pre-trained BEATs model near-far microphone spatial statistics
40 Zheng_HFUUAI_task2_2 ZhengHFUUAI2026 kNN, score fusion BEATs LoRA embeddings, EAT embeddings, spatial statistics noise mixing, random gain, time shift weighted average of normalized anomaly scores BEATs, EAT 4 pre-trained BEATs and EAT models near-far microphone spatial statistics
25 Zheng_HFUUAI_task2_3 ZhengHFUUAI2026 kNN, score fusion BEATs LoRA embeddings, spatial statistics noise mixing, random gain, time shift weighted average of normalized anomaly scores BEATs 2 pre-trained BEATs model near-far microphone spatial statistics
123 Zheng_HFUUAI_task2_4 ZhengHFUUAI2026 kNN, score fusion BEATs LoRA embeddings, EAT embeddings, spatial statistics noise mixing, random gain, time shift weighted average of normalized anomaly scores BEATs, EAT 3 pre-trained BEATs and EAT models near-far microphone spatial statistics
28 Zhou_HFUUDS_task2_1 ZhouHFUUDS2026 ArcFace, EAT, kNN, LoRA, memory bank EAT token maps, log-mel energies supplemental mixup thresholding on anomaly score EAT large pretrained on AudioSet-2M pretrained model Mel spectrogram
22 Zhou_HFUUDS_task2_2 ZhouHFUUDS2026 ArcFace, EAT, kNN, LoRA, memory bank EAT token maps, log-mel energies supplemental mixup thresholding on anomaly score EAT large pretrained on AudioSet-2M pretrained model Mel spectrogram
26 Zhou_HFUUDS_task2_3 ZhouHFUUDS2026 ArcFace, EAT, kNN, LoRA, memory bank EAT token maps, log-mel energies supplemental mixup thresholding on anomaly score EAT large pretrained on AudioSet-2M pretrained model Mel spectrogram
20 Zhou_HFUUDS_task2_4 ZhouHFUUDS2026 ArcFace, EAT, kNN, LoRA, memory bank EAT token maps, log-mel energies supplemental mixup thresholding on anomaly score EAT large pretrained on AudioSet-2M pretrained model Mel spectrogram
95 Zhu_FDA_task2_1 ZhuFDA2026 EAT LoRA, kNN, Mahalanobis distance, score fusion log-mel energies and statistical audio features frequency masking, gain scaling, time masking, time shifting fixed score fusion and train-normal quantile threshold EAT 2 pre-trained EAT model two-channel near/far feature extraction
115 Zhu_FDA_task2_2 ZhuFDA2026 EAT LoRA, kNN, Mahalanobis distance, score fusion log-mel energies and statistical audio features frequency masking, gain scaling, time masking, time shifting fixed score fusion and train-normal quantile threshold EAT 2 pre-trained EAT model two-channel near/far feature extraction
100 Zhu_FDA_task2_3 ZhuFDA2026 EAT LoRA, kNN, Mahalanobis distance, score fusion log-mel energies and statistical audio features frequency masking, gain scaling, time masking, time shifting fixed score fusion and train-normal quantile threshold EAT 2 pre-trained EAT model two-channel near/far feature extraction
94 Zhu_FDA_task2_4 ZhuFDA2026 EAT LoRA, kNN, Mahalanobis distance, score fusion log-mel energies and statistical audio features frequency masking, gain scaling, time masking, time shifting fixed score fusion and train-normal quantile threshold EAT 2 pre-trained EAT model two-channel near/far feature extraction
27 Huang_QWS_task2_1 HuangQWS2026 log-mel energies, raw waveform SpecAugment weighted average BEATs
146 Wang_Liu_SuzhouDongyuan_task2_1 WangLiu2026 LoRA-adapted pre-trained audio encoders, ArcFace representation learning, Mahalanobis distance, kNN cosine distance, score-level ensemble raw waveform input to pre-trained BEATs and AudioMAE encoders channel dropout, channel gain perturbation, channel swap robust-z weighted average, BEATs weight 0.85, AudioMAE weight 0.15 AudioMAE, BEATs 2 pre-trained models fixed-length waveform crop and padding, inference-time left-channel duplication
163 Wang_Liu_SuzhouDongyuan_task2_2 WangLiu2026 LoRA-adapted pre-trained audio encoders, ArcFace representation learning, Mahalanobis distance, kNN cosine distance, score-level ensemble raw waveform input to pre-trained BEATs and AudioMAE encoders channel dropout, channel gain perturbation, channel swap robust-z weighted average, BEATs weight 0.85, AudioMAE weight 0.15 AudioMAE, BEATs 2 pre-trained models fixed-length waveform crop and padding, original dual-channel gated inference
168 Wang_Liu_SuzhouDongyuan_task2_3 WangLiu2026 LoRA-adapted pre-trained audio encoders, ArcFace representation learning, Mahalanobis distance, kNN cosine distance, score-level ensemble raw waveform input to pre-trained BEATs and AudioMAE encoders channel dropout, channel gain perturbation, channel swap robust-z weighted average, BEATs weight 0.90, AudioMAE weight 0.10 AudioMAE, BEATs 2 pre-trained models fixed-length waveform crop and padding, inference-time left-channel duplication
152 Wang_Liu_SuzhouDongyuan_task2_4 WangLiu2026 LoRA-adapted pre-trained BEATs encoder, ArcFace representation learning, Mahalanobis distance raw waveform input to pre-trained BEATs encoder channel dropout, channel gain perturbation, channel swap BEATs pre-trained model fixed-length waveform crop and padding, inference-time left-channel duplication
23 Qian_nivic_task2_1 Qiannivic2026 kNN log-mel energies mixup Spectral Subtraction
8 Qian_nivic_task2_2 Qiannivic2026 kNN log-mel energies mixup Spectral Subtraction
42 Qian_nivic_task2_3 Qiannivic2026 kNN log-mel energies mixup Spectral Subtraction
12 Qian_nivic_task2_4 Qiannivic2026 kNN log-mel energies mixup Spectral Subtraction
119 Jeong_Medisensing_task2_1 JeongMedisensing2026 EAT (frozen) + LoRA, supervised contrastive learning, normalizing flow, Mahalanobis (on EAT and raw log-mel), CNN, ridge regression, cosine kNN, ensemble log-mel energies 90th-percentile threshold of training scores EAT (pre-trained on AudioSet) 7 pre-trained model learned cross-channel representation (ILD/IPD CNN) + cross-channel predictive residual + RTF spatial features + raw log-mel energies
135 Jeong_Medisensing_task2_2 JeongMedisensing2026 EAT (frozen) + LoRA, supervised contrastive learning, normalizing flow, Mahalanobis distance, CNN, ridge regression, ensemble log-mel energies 90th-percentile threshold of training scores EAT (pre-trained on AudioSet) 5 pre-trained model learned cross-channel representation (ILD/IPD CNN, emphasized) + cross-channel predictive residual + RTF spatial features
125 Jeong_Medisensing_task2_3 JeongMedisensing2026 EAT (frozen) + LoRA, supervised contrastive learning, normalizing flow, Mahalanobis (on EAT and raw log-mel), CNN, ridge regression, ensemble log-mel energies 90th-percentile threshold of training scores EAT (pre-trained on AudioSet) 6 pre-trained model RTF spatial features (emphasized) + learned cross-channel representation (ILD/IPD CNN) + cross-channel predictive residual + raw log-mel energies
106 Jeong_Medisensing_task2_4 JeongMedisensing2026 EAT (frozen) + LoRA supervised contrastive learning, normalizing flow, Mahalanobis on EAT embeddings and on raw log-mel features, ensemble log-mel energies 90th-percentile threshold of training scores EAT (pre-trained on AudioSet) 3 pre-trained model raw log-mel energies (Selective Mahalanobis)
37 Zhou_XAUAT_task2_1 ZhouXAUAT2026 Conformer, kNN log-mel energies mixup, SpecAugment
101 Kim_CAU_task2_1 KimCAU2026 k-means, kNN log-mel energies EAT pre-trained model CMSN
66 Kim_CAU_task2_2 KimCAU2026 k-means, kNN log-mel energies EAT pre-trained model CMSN
86 Kim_CAU_task2_3 KimCAU2026 k-means, kNN log-mel energies EAT pre-trained model CMSN
81 Kim_CAU_task2_4 KimCAU2026 ensemble, k-means, kNN log-mel energies weighted average EAT 3 pre-trained model CMSN
165 Zeng_BUCT_task2_1 ZengBUCT2026 BEATs asp arcface log-mel filterbank mixup, SpecAugment BEATs
171 Zeng_BUCT_task2_2 ZengBUCT2026 CNN log-mel energies mixup average
172 Zeng_BUCT_task2_3 ZengBUCT2026 BEATs asp arcface log-mel filterbank mixup, SpecAugment BEATs
162 Zeng_BUCT_task2_4 ZengBUCT2026 BEATs asp arcface log-mel filterbank mixup, SpecAugment BEATs
43 Yang_NJU_task2_1 YangNJU2026 attentive statistics pooling, BEATs, kNN, LoRA fine-tuning, neural enhancement, teacher-student knowledge distillation, two-channel enhanced-noisy fusion log-mel energies, raw waveform SpecAugment BEATs pre-trained model neural enhancement front-end
57 Yang_NJU_task2_2 YangNJU2026 attentive statistics pooling, BEATs, kNN, LoRA fine-tuning, neural enhancement, teacher-student knowledge distillation, two-channel enhanced-noisy fusion log-mel energies, raw waveform SpecAugment BEATs pre-trained model neural enhancement front-end
73 Kim_KATECH_task2_1 KimKATECH2026 AE log-mel energies average 10 noise-aware far-mic reference denoising; per-band adaptive spectral subtraction (AdaSub); per-domain-centered reconstruction-energy term
80 Kim_KATECH_task2_2 KimKATECH2026 AE log-mel energies average 10 noise-aware far-mic reference denoising; fixed-coefficient spectral subtraction (RefSub, alpha=1.5 beta=0.10); per-domain-centered reconstruction-energy term
61 Kim_KATECH_task2_3 KimKATECH2026 AE log-mel energies average 20 noise-aware far-mic reference denoising; per-clip blend of AdaSub and inter-channel coherence-gate (CohGate); per-domain-centered reconstruction-energy term
77 Kim_KATECH_task2_4 KimKATECH2026 AE log-mel energies average 10 noise-aware far-mic reference denoising; cross front-end (per-band adaptive spectral subtraction [specsubada] and fixed-coefficient spectral subtraction [refsub]), dev-Omega-selected; per-domain-centered reconstruction-energy term
79 Guan_GISP@HEU_task2_1 GuanGISP@HEU2026 kNN log-mel energies, raw waveform average BEATs, CED, EAT pre-trained model
38 Krag_AAU_task2_1 KragAAU2026 AE, kNN log-mel energies, spectrogram gaussian white noise, time shifting minimum BEATs, SSLAM 2
104 Krag_AAU_task2_2 KragAAU2026 AE, GMM, kNN log-mel energies, spectrogram gaussian white noise, time shifting minimum BEATs, SSLAM 2
116 Krag_AAU_task2_3 KragAAU2026 AE, kNN log-mel energies, spectrogram gaussian white noise, time shifting minimum BEATs, SSLAM 3
19 Krag_AAU_task2_4 KragAAU2026 AE, GMM, kNN log-mel energies, spectrogram gaussian white noise, time shifting minimum BEATs, SSLAM 3
52 Moradi_JKU_task2_1 MoradiJKU2026 ensemble, kNN, local-density KNN, PCA-Mahalanobis frozen BEATs embeddings, frozen EAT-base embeddings shiftall embedding mean-shift augmentation weighted average, threshold at zero BEATs, EAT-base 4 pre-trained model embeddings stereo Wiener denoising, near-far channel difference
60 Moradi_JKU_task2_2 MoradiJKU2026 ensemble, kNN, local-density KNN, PCA-Mahalanobis frozen BEATs embeddings, frozen EAT-base embeddings shiftall embedding mean-shift augmentation weighted average, threshold at zero BEATs, EAT-base 4 pre-trained model embeddings stereo Wiener denoising, near-far channel difference



Technical reports

MULTI-ENCODER FUSION WITH LORA AND RANK NORMALIZA TION FOR FIRST-SHOT ANOMALOUS SOUND DETECTION

Nehemiah Balozi, Sungkyunkwan
Department of Intelligent Robotics Engineering, Sungkyunkwan University, Suwon, South Korea and Department of Mechanical Engineering, Sungkyunkwan University, Suwon, South Korea

Abstract

We present a multi-encoder fusion system for noise-aware unsupervised anomalous sound detection. Six pretrained audio encoders are combined: frozen CLAP, WavLM-Base, and AST with trainable projection heads, plus BEATs adapted via Low-Rank Adaptation (LoRA) at two ranks (r=16, r=32) and CLAP adapted via LoRA. Adapter layers and projection heads are trained on normal data using InfoNCE contrastive loss. Anomaly scores are computed via k-nearest neighbor distance in 256-dimensional embedding space, normalized using per-machine per-domain rank normalization, and fused via weighted score averaging. Key findings include: LoRA adaptation of BEATs and CLAP improved performance over frozen encoders in our development experiments, and per-(machine, domain) rank normalization improved development-set performance and was adopted in all submitted systems. On the DCASE 2026 Task 2 development dataset, our best six-stream system achieves a development-set macro(3) of 60.06% across the seven development machine types.

System characteristics
Classifier ensemble, kNN
System complexity 365000000
Acoustic features raw waveform
Data augmentation MemMixup
Decision making weighted average
System embeddings AST (all12), BEATs (LoRA r=16), BEATs (LoRA r=32), CLAP, CLAP (LoRA, all12), WavLM, AST, BEATs (LoRA r=16), BEATs (LoRA r=32), CLAP, CLAP (LoRA), WavLM, BEATs (LoRA r=16), BEATs (LoRA r=32), CLAP, CLAP (LoRA), WavLM, CLAP, WavLM, BEATs (LoRA r=16, all12), BEATs (LoRA r=32, all12), CLAP (LoRA, all12), AST (all12)
Subsystem count 5, 6
External data usage pre-trained model
PDF

A FIRST-SHOT UNSUPERVISED ANOMALOUS SOUND DETECTION SYSTEM FOR DCASE2026 TASK 2

Peiwei Chang, Yuelan Cheng, Yongqiang Chen, Philip J.B. Jackson, Wenwu Wang
Centre for Vision, Speech and Signal Processing (CVSSP), University of Surrey, Guildford, UK

Abstract

We present our submissions to Detection and Classification of Acoustic Scenes and Events (DCASE) 2026 Challenge Task 2 (Noise-aware Unsupervised Anomalous Sound Detection for Machine Condition Monitoring). We combine up to ten complementary anomaly signals using a per-machine, label-free adaptive router. Each component is weighted by the Kolmogorov–Smirnov distance between its test-clip and train-clip score distributions, requiring no anomaly labels and no cross-machine fitting. The signals span frozen AudioSet self-supervised embeddings (M2D, M2DCLAP, BEATs, EAT) scored by cosine k-NN, classical conditionmonitoring and inter-channel spatial features, a domain-adapted M2D obtained by further masked-modeling pre-training on machine audio (broadband and 4–7.9 kHz band), and a two-microphone spatial noise-cancellation channel that removes the coherent factory noise before re-embedding. Our system uses per-clip weightedmaximum fusion, treating anomaly detection as an existential event over components. On the development set, our submitted system achieves a source-domain AUC of 76.3% a target-domain AUC of 68.0%, and a pAUC of 56.9%.

System characteristics
Classifier Mahalanobis distance, ensemble, k-NN (cosine)
System complexity 180691327, 440525701, 7850
Acoustic features condition-monitoring indicators, inter-channel spatial features, log-mel energies, raw waveform
Decision making weighted average, weighted maximum
System embeddings BEATs, EAT, M2D, M2D-CLAP, BEATs, EAT, domain-adapted M2D (further pre-trained on machine audio)
Subsystem count 10, 2, 7
External data usage pre-trained model
Front end system 2-mic Wiener spatial noise cancellation, band-pass filtering (4-7.9kHz)
PDF

Description and Discussion on DCASE 2026 Challenge Task 2: Noise-aware Unsupervised Anomalous Sound Detection for Machine Condition Monitoring

Tomoya Nishida, Noboru Harada, Daiki Takeuchi, Daisuke Niizumi, Keisuke Imoto, Kota Dohi, Harsh Purohit, Takashi Endo, Yohei Kawaguchi
Research and Development Group, Hitachi, Ltd., Tokyo, Japan and NTT Corporation, Kanagawa, Japan

Abstract

This paper presents an overview of DCASE 2026 Challenge Task 2, titled “Noise-aware unsupervised anomalous sound detection (UASD) for machine condition monitoring.” The task aims to advance noise-robust anomalous sound detection for machine condition monitoring under the unsupervised setting, where only normal machine sounds are available for training. Reliable detection under noisy conditions is crucial for practical deployment, but previous DCASE Task 2 settings provided limited information about environmental noise, potentially limiting UASD performance in highly noisy situations. To address this limitation, DCASE 2026 allows participants to exploit two-channel audio samples simultaneously captured at locations near and far from the target machine. Since the distant microphone is expected to contain relatively stronger environmental noise and weaker direct machine sounds, it may help distinguish environmental noise components from the target machine sounds. After the challenge submission deadline, challenge results and an analysis of the submitted systems will be added.

System characteristics
Classifier AE
System complexity 269992
Acoustic features log-mel energies
PDF

THE WISTLAB SYSTEM FOR DCASE 2026 TASK 2: FINE-GRAINED SCORING FOR NOISE-A W ARE MACHINE SOUND ANOMALY DETECTION

Pingyi Fan, Anbai Jiang, Shuwei Zhang, Lvxin Xu, Wenrui Liang, Tianyu Liu, Xinhu Zheng, Junjie Li, Wei-Qiang Zhang, Cheng Lu, Yanmin Qian, Xie Chen, Jia Liu
Department of Electronic Engineering, Tsinghua University, Beijing, China and Department of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai, China and Schools of Economy, North China Electric Power University, Beijing, China

Abstract

This report outlines the WISTLAB submission to DCASE 2026 Challenge Task 2 on noise-aware machine sound anomaly detection. The system uses task-adapted BEATs representations together with sub-band scoring, fine-grained localized scoring, and local-density normalization. The sub-band component extends AdaBEAM with a learned attentive-pooling view, while the fine-grained component retains local time-frequency evidence for normal-reference scoring. We submit one single scoring system and three score-level fusion systems, with a best development-set harmonic mean of 67.20%.

System characteristics
Classifier ensemble, pre-trained models
System complexity 180000000, 540000000, 630000000, 90000000
Acoustic features STFT, log-mel filterbank
Data augmentation SpecAugment
Decision making Bayesian optimization
Subsystem count 2, 6, 7
External data usage BEATs
PDF

THE MERL SYSTEMS FOR DCASE 2026 CHALLENGE TASK 2

Takuya Fujimura, Gordon Wichern, Yoshiki Masuyama, Christoph Boeddeker, Kohei Saijo, Julius Richter, Takahiro Edo, Jonathan Le Roux, 1Mitsubishi Electric
Mitsubishi Electric Research Laboratories, Cambridge, USA

Abstract

In this report, we present our anomalous sound detection (ASD) systems for DCASE 2026 Challenge Task 2. Our approach introduces noise-aware audio self-supervised learning (NA-SSL) to leverage two-channel recordings, in which one microphone is used to capture noise. NA-SSL models are trained to extract clean SSL representations from two-channel noisy signals simulated on external datasets. Then, we perform ASD in the extracted denoised feature space. To further improve performance, we perform discriminative fine-tuning with attributes and pseudo labels. Furthermore, for anomaly score calculation, we employ several recent techniques: score rescaling, frequency-wise memory bank construction, and deviation-based pooling. Our final ensemble system has achieved 66.20% in the official scores calculated as a harmonic mean of the area under the curve (AUC) and partial AUC (p= 0.1) over all machine types and domains in the development set.

System characteristics
System complexity 1106841600, 1125553488, 1582854592, 565054800
Acoustic features log-mel energies
Data augmentation mixup
Decision making average
System embeddings BEATs, BEATs, EAT, BEATs, EAT, EAT large, EAT
Subsystem count 3, 6, 9
External data usage pre-training
PDF

FIRST-SHOT ANOMALOUS SOUND DETECTION WITH FROZEN GENERAL AUDIO ENCODERS AND DISTANCE-BASED BACK-ENDS

Rene Glitza Luca Becker and Rainer Martin, Ruhr-
Institute of Communication Acoustics, Ruhr University Bochum, Bochum, Germany

Abstract

We present a training-free approach to first-shot unsupervised anomalous sound detection in the noise-aware, two-channel setting of DCASE 2026 Task 2. Instead of training or fine-tuning on the challenge data, we pass each recording through a frozen, general-purpose audio encoder, fit a simple reference model on the few normal clips available for each machine, and score a test clip by how far it sits from that reference, so adapting to a new machine or domain only means replacing the reference set. We vary two choices: the embedding source, where we put the industrialsignal foundation model FISHER against an Audio Spectrogram Transformer (AST) that we self-pre-train on AudioSet, and the scoring back-end, a non-parametrick-nearest-neighbor (kNN) memory bank against a per-machine Gaussian mixture model (GMM), both with domainz-score calibration. On the development set the frozen FISHER encoder with akNN back-end reaches the best official score (Ω = 58.23%) and is the only configuration to beat both the autoencoder and selective-Mahalanobis baselines, which shows that frozen embeddings with a simple estimator can compete with task-specific reconstruction models in the first-shot regime.

System characteristics
Classifier GMM, kNN
System complexity 21393408, 85254144
Acoustic features log-mel energies, spectrogram
Decision making threshold
System embeddings FISHER
PDF

GISP@HEU’S SUBMISSION FOR DCASE 2026 TASK 2: TRAINING-FREE ANOMALOUS SOUND DETECTION WITH ADAPTIVE LA YER SELECTION

Tong Ye, Yao Xiao, Wenbo Wang, Qiaoxi Zhu, Jian Guan, 1Group of Intelligent Signal Processing, Harbin Engineering
Harbin Engineering University, Harbin, China and University of Technology Sydney, Ultimo, Australia

Abstract

This report presents our submission to Task 2 of the DCASE 2026 Challenge, which focuses on developing noise-robust unsupervised anomalous sound detection (UASD) systems for first-shot machine condition monitoring under domain-shift conditions. To address this challenge, we propose a fully training-free system consisting of three frozen audio pre-trained models (i.e., BEATs, CED, and EAT) together with an adaptive layer selection strategy to obtain generalizable and robust machine sound representations for anomalous detection. Experimental results on the DCASE 2026 Task 2 development set demonstrate the effectiveness of the proposed training-free system, with the harmonic-mean AUC and harmonic-mean pAUC reaching 53.7% and 59.4%, respectively.

System characteristics
Classifier kNN
System complexity 300000000
Acoustic features log-mel energies, raw waveform
Decision making average
System embeddings BEATs, CED, EAT
External data usage pre-trained model
PDF

GENREP WITH MULTI-BRANCH DUAL-CHANNEL INPUTS AND ADAPTIVE POOLING FOR TRAINING-FREE ANOMALOUS SOUND DETECTION

Chaoyong Huang, Xiangyu Jing Yuandong Luo, Hongqing Liu, 1School of Communications and Information Engineering, Chongqing
School of Communications and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing, China

Abstract

This technical report describes our submission to the DCASE 2026 Challenge Task 2 on first-shot unsupervised anomalous sound detection (ASD) under domain shift. Our system builds on the GenRep framework proposed by Saengthong and Shinozaki, which performs training-free ASD using frozen embeddings from largescale pre-trained audio encoders withk-nearest neighbor (kNN) scoring and domain-wise score normalization. We extend GenRep in three directions motivated by the dual-channel (near/far) audio setup newly introduced in DCASE 2026. First, we construct four input methods, including single-channel, far-weighted concatenation, inter-channel absolute difference, and STFT-domain soft-mask enhancement, each with parameter variants yielding seven configurations. Second, we incorporate embedding preprocessing and replace default mean pooling with adaptive temporal pooling strategies including generalized mean (GeM) pooling, relative deviation pooling (RDP), and hybrid RDP+GeM, following Wilkinghoff et al. Third, we perform per-encoder search over backend configurations and fuse complementary high-performing candidates viaZscore-aligned weighted score averaging. Our systems substantially outperform the official baselines on the development set.

System characteristics
Classifier frozen pre-trained audio encoder, frozen pre-trained audio encoders, kNN, score fusion
System complexity 183000000, 278000000, 371000000, 88000000
Acoustic features raw waveform
Decision making train-normal z-score alignment, weighted average
System embeddings Dasheng, EAT-AS2M, EAT-large, OpenL3, SSLAM, Dasheng, EAT-large, OpenL3, Dasheng, EAT-large, SSLAM, EAT-large
Subsystem count 3, 5
External data usage pre-trained audio embeddings
Front end system STFT soft mask
PDF

ANOMALOUS SOUND DETECTION BASED ON NOISE REFERENCE ENHANCEMENT AND ADAPTIVE SCORE FUSION

Kaixing Ding, Da Huang, Aoyu Liu, Quectel Wireless Solutions Co
Quectel Wireless Solutions Co., Ltd, Hefei, China

Abstract

This technical report presents the competition system designed by our team for the DCASE 2026 Anomalous Sound Detection (ASD) Challenge. We propose an anomalous sound detection system based on noise reference enhancement and adaptive score fusion. The system first estimates the noise power spectrum using the reference channel, computes the frequency -domain gain combined with the spectrum of the target channel, and obtains the enhanced audio. Then, complementary acoustic features are extracted, and a regular ized statistical distance model is constructed based on normal samples. The multi-feature scores are adaptively fused according to the domain distance between the feature distributions of the source domain and target domain, and geometric product is utilized to impose consistency constraints on anomaly scores, so that samples judged as anomalous from multiple perspectives obtain higher anomaly scores. The system only requires normal audio clips from the development and evaluation datasets, with no need for synthetic anomalous samples. Test res ults on 7 types of development datasets show that the harmonic mean of AUC and pAUC of the proposed system reaches 0.6544, which surpasses the official baseline , and can meet the requirements of unsupervised anomaly detection in noisy industrial scenarios.

System characteristics
System complexity 90000000
Acoustic features log-mel energies, raw waveform
Data augmentation SpecAugment
Decision making weighted average
System embeddings BEATs
PDF

NOISE-A W ARE UNSUPERVISED ANOMALOUS SOUND DETECTION WITH PRE-TRAINED AUDIO ENCODERS AND LORA ADAPTATION

Huan Yu, Yanjin Li, Xuanting Fan, Pan Li, Zongmu Lin, Zhongxin Bai, Gongping Huang
Wuhan University, Wuhan, China and University of International Business and Economics, Beijing, China and Harbin Engineering University, Harbin, China

Abstract

This technical report presents our submission to the DCASE 2026 Challenge Task 2. We adopt a two-stage framework consisting of a front-end feature extraction module and a back-end anomaly scoring module. Specifically, we extract representations using four pretrained audio encoders-ATST-Frame, SSLAM, EAT-large, and BEATs-together with their LoRA-adapted variants. Features are extracted from multiple intermediate layers to enhance representational diversity. To improve robustness, the extracted features are further whitened before anomaly scoring. We investigate three complementary scoring strategies, including local-density knearest neighbors, relative Mahalanobis distance, and a hybrid GMM–cosine–kNN detector. The final anomaly score is obtained by first fusing scores across layers within each encoder and then aggregating them across different encoders. By ensembling different front-end representations and back-end scoring methods, we construct four systems for submission. On the development set, these systems achieve official scores of 66.65%, 65.58%, 64.02%, and 61.35%, respectively, consistently outperforming the official baseline of 57.66%.

System characteristics
Classifier GMM, LoRA, Mahalanobis distance, cosine similarity, ensemble, kNN, local density normalization
System complexity 1200000000, 1461884160, 2439509248, 312000000
Acoustic features SSL transformer layer embeddings, log-mel energies
Data augmentation mixup, waveform perturbation
Decision making average, gamma-fit threshold on train normal scores (90th percentile), rank-normalized weighted average, rank_mean (branch fusion)
System embeddings ATST-Frame, BEATs, EAT-large, SSLAM, BEATs, EAT-large, SSLAM, BEATs_ft1, CED-tiny, EAT-large, M2D-CLAP, SSLAM
Subsystem count 15, 5, 8, 9
External data usage pre-trained model, pre-trained model, LoRA fine-tuning, pre-trained model, embeddings
Front end system inter-channel SNR profiling, stereo channel alignment, SNR-aware enhancement, layer EMA smoothing
PDF

Noise-Aware Far-Conditioned Masked Spectrogram Modeling with DSP Feature Fusion DCASE 2026 Challenge Task 2 -

Iljoo Jeong, Korea Electronics Technology
Korea Electronics Technology Institute (KETI), Seongnam, South Korea

Abstract

We address first-shot, noise-aware anomalous sound detection under source and target domain shift, using the synchronized Near and Far (close and distant) stereo recordings, and we train the entire system from scratch without any pretrained models or external data. Our design is motivated by an invariance trap that we observe in this setting: representations learned from scratch with discriminative or denoising objectives, such as machine-ID classification, pseudo-cluster metric learning, and far-to-near denoising, collapse fault sensitivity and reach near-chance detection, because one-class training provides no gradient toward fault-relevant directions. To avoid this, we combine two complementary branches. First, we preserve fault-relevant time-frequency structure with windowed masked spectrogram modeling (MSM), namely a convolutional encoder-decoder that is trained per machine to reconstruct 60%-masked 2.5 s log-mel patches from a bottleneck embedding; the embedding itself, rather than its reconstruction residual, is used for scoring. Second, we inject fault-axis sensitivity that learning alone does not discover, using non-neural DSP descriptors computed directly from each Near and Far window, i.e., a far-context ridge-residual descriptor and a temporal modulation descriptor. From normal Near and Far pairs we estimate a Far-correlated shared nuisance subspace, and we form a Far-conditioned MSM feature that attenuates this shared operational and environmental variation while preserving Near-specific variation. The three blocks are block-balanced, concatenated, and scored by a single domain-conditional kNN density estimator with target-domain shrinkage, and the window scores are aggregated by a top-10% mean into one anomaly score per file. A single scalar that controls the Far-conditioning is selected on the development machines and fixed for evaluation, so that no per-machine tuning is applied and no anomaly labels are used during training or model selection. We use the DCASE 2026 Task 2 dataset and challenge protocol [1]. The dataset and task build on prior anomalous-sound-detection resources, namely ToyADMOS2 [2] and MIMII DG [3], and we follow the first-shot domain-generalization baseline [4].

System characteristics
Classifier kNN density, masked spectrogram autoencoder (CNN)
System complexity 1345121
Acoustic features far-context ridge-residual descriptor, log-mel energies, temporal modulation descriptor
Data augmentation spectrogram patch masking
PDF

CROSS-CHANNEL AND RA W-SPECTRAL DETECTOR ENSEMBLES FOR FIRST-SHOT NOISE-A W ARE ANOMALOUS SOUND DETECTION

Seunggyu Jeong, Seong-Eun Kim
Medisensing & Seoul National University of Science and Technology (SeoulTech), Seoul, Korea

Abstract

We describe our submission to DCASE 2026 Challenge Task 2, first-shot unsupervised anomalous sound detection (ASD) for machine condition monitoring under a new noise-aware, two-channel (near and far) recording setup. Our system combines a selfsupervised representation, an EAT audio transformer adapted with LoRA and trained with a composite supervised-contrastive objective, source-target alignment, and pseudo-attribute clustering, with an ensemble of orthogonal anomaly detectors. On top of the density core (Mahalanobis distance and a per-machine normalizing flow) we add two cross-channel detectors that use the second channel through complementary mechanisms: a learned cross-channel representation (LCCR) that classifies inter-channel level and phase maps, and a cross-channel predictive consistency score (C1) whose residual flags anomalies that the discriminative branch misses. A third detector scores raw log-mel energies with a selective Mahalanobis distance and is strong on the machines where the deep embedding is weak. All scores are combined by a domain-agnostic rank fusion, so the pipeline can be applied to the unlabelled evaluation set. On the development set the primary system reaches a harmonic-meanΩof62.95, above the official baseline (57.7). We submit four systems that trade off reliance on the deep embedding against the raw-spectral detector.

System characteristics
Classifier CNN, EAT (frozen) + LoRA, EAT (frozen) + LoRA supervised contrastive learning, Mahalanobis (on EAT and raw log-mel), Mahalanobis distance, Mahalanobis on EAT embeddings and on raw log-mel features, cosine kNN, ensemble, normalizing flow, ridge regression, supervised contrastive learning
System complexity 92276000
Acoustic features log-mel energies
Decision making 90th-percentile threshold of training scores
System embeddings EAT (pre-trained on AudioSet)
Subsystem count 3, 5, 6, 7
External data usage pre-trained model
Front end system RTF spatial features (emphasized) + learned cross-channel representation (ILD/IPD CNN) + cross-channel predictive residual + raw log-mel energies, learned cross-channel representation (ILD/IPD CNN) + cross-channel predictive residual + RTF spatial features + raw log-mel energies, learned cross-channel representation (ILD/IPD CNN, emphasized) + cross-channel predictive residual + RTF spatial features, raw log-mel energies (Selective Mahalanobis)
PDF

AITHU SUBMISSION FOR DCASE 2026 TASK 2: ROBUST MACHINE-WISE SCORING WITH BEA TS REPRESENTA TIONS

Anbai Jiang, Xinhu Zheng, Wenrui Liang, Shuwei Zhang, Tianyu Liu, Jia Liu, Pingyi Fan, Wei-Qiang Zhang, Cheng Lu, Xie Chen, Yanmin Qian
Department of Electronic Engineering, Tsinghua University, Beijing, China and Department of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai, China and Huakong AI Plus Company Limited, Beijing, China and Schools of Economy, North China Electric Power University, Beijing, China

Abstract

This report describes the AITHU submission to DCASE 2026 Challenge Task 2 on noise-aware anomalous sound detection. The central component of our system is a score-level fusion that exploits the complementarity of multiple independently trained systems. All systems are built on BEATs-based audio representations with distance-based, primarily Mahalanobis, anomaly scoring, and BEATs is the only external pre-trained model, so the diversity required for an effective fusion comes from heterogeneous scoring branches, training seeds, and generative augmentation rather than from additional backbones. The branches are combined by Bayesian-optimized score-level selection. Four systems are submitted, including one single scoring system and three ensemble systems. The best submitted system achieves a development-set harmonic mean of 68.20%.

System characteristics
Classifier Mahalanobis distance, ensemble, pre-trained models
System complexity 2070000000, 630000000, 90000000, 900000000
Acoustic features STFT, log-mel filterbank
Data augmentation SpecAugment
Decision making Bayesian optimization
Subsystem count 10, 3, 7
External data usage BEATs
PDF

DUAL-CHANNEL CROSS-A TTENTION EMBEDDINGS FOR NOISE-A W ARE UNSUPERVISED ANOMALOUS SOUND DETECTION

Ziran Jiang, Rong Han
KY, Beijing, China

Abstract

This technical report describes our submission to DCASE 2026 Challenge Task 2: first-shot unsupervised anomalous sound detection for machine condition monitoring. The proposed system learns compact representations from normal training recordings using a dual-channel convolutional encoder followed by cross-channel attention. The near- and far-microphone log-mel spectrograms are encoded separately, exchanged through bidirectional cross-attention layers, and averaged into a single normalized embedding. Anomaly scores are computed without anomalous training data by fitting Gaussian models to the training embeddings and using the minimum Mahalanobis distance as the score. For the official evaluation, we submit four systems that differ in model resolution and checkpoint-ensemble strategy: two 128-bin mel ensembles and two 256-bin mel ensembles.

System characteristics
Classifier CNN, cross-attention
System complexity 4843392
Acoustic features log-mel energies
Data augmentation SpecAugment
Decision making average
Subsystem count 10, 20
PDF

A TRAIN-NORMAL PROFILE ENSEMBLE FOR NOISE-AWARE UNSUPERVISED ANOMALOUS SOUND DETECTION

Hisataka Kajita , Hisataka Kajita
Independent, Independent Researcher, Japan

Abstract

This report describes our submission to DCASE 2026 Challenge Task 2 [1], which targets first -shot unsupervised anomalous sound detection for machine condition monitoring under noise and domain shift. The submitted systems avoid using evaluation-test labels, source or target domain labels, and test attributes. Each evaluation machine is modeled only from its available normal training recordings. We build multiple interpretable anomaly-score components from robust acoustic statistics, two - channel near/far relations, percentile ranks, high -tail and low-tail departures, local -neighbor distances, and reconstruction-style residuals. Four submitted outputs differ only in their train -normal score aggregation rule: domain -conditioned consensus, top -tail rescue, two-sided rank scoring, and agreement -balanced consensus. Development experiments showed that simple single -model distances were insufficient, motivating multi -axis score families and conservative rank -shaping variants. For the final evaluation data, ground-truth labels are hidden, so official test performance is computed by the challenge organizers.

System characteristics
Classifier Mahalanobis distance, PCA residual scoring, kNN, rank-based ensemble, robust statistics
System complexity 0
Acoustic features clip-level acoustic statistics, periodicity descriptors, spectral descriptors, two-channel near/far relation descriptors
Decision making train-normal 99th percentile threshold
Subsystem count 54
Front end system robust clip-level normalization
PDF

AISTAT LAB SYSTEM FOR DCASE 2026 TASK 2: NEAR-ANCHOR REPRESENTATION LEARNING WITH SELECTIVE FUSION

Min Jun Kim Jeong Ho Seo Sang Chun Park Seung Woo Sin Changwon Lim, 1Department of Applied Statistics, Chung-Ang
Department of Applied Statistics, Chung-Ang University, Seoul, Korea and Department of Mathematics, Chung-Ang University, Seoul, Korea

Abstract

This report describes the AISTAT LAB submission to DCASE 2026 Task 2 for noise-aware first-shot anomalous sound detection with synchronized near and far channel recordings under source and target domain shift. We propose a two-stage framework based on Efficient Audio Transformer encoders with channel-wise CMSN, attentive statistics pooling, and near-anchor DualPoolFusion. The fused representation is trained with Angular–Euclidean Compactness (AEC) using machine–domain–attribute labels. For inference, we combine KMeans-based cosine scoring, Relative Reference Distance Score (RRDS), and sample-wise adaptive RRDS weighting. The weighted ensemble achieves the best official score of 64.86% on the development-test set.

System characteristics
Classifier ensemble, k-means, kNN
System complexity 249000000, 747000000
Acoustic features log-mel energies
Decision making weighted average
System embeddings EAT
Subsystem count 3
External data usage pre-trained model
Front end system CMSN
PDF

Noise-Aware Reference Denoising for First-Shot Anomalous Sound Detection

Nam Kyun Kim, Automotive Electronics R&D
Automotive Electronics R&D Center, Korea Automotive Technology Institute (KATECH), Gwangju, Republic of Korea

Abstract

This report addresses the noise-aware first-shot unsupervised anomalous sound detection (ASD) task of the DCASE 2026 Challenge Task 2, where each clip has a near and a far microphone but the official baseline scores only the near channel. The far microphone is exploited as a noise reference: a per-machine noise-transfer function is built from each channel’s minimum-statistics noise floor rather than the full far spectrum, so that machine sound leaking into the far channel is preserved. The near channel is then denoised by floored spectral subtraction before a reconstruction autoencoder with a Mahalanobis score, and a per-band adaptive variant that over-subtracts where the local SNR is low performs best. Every statistic is computed from training-normal clips only, and all detectors are reported as the mean ±std over ten autoencoder seeds. The noise-aware denoising robustly lifts the official development score from the baseline 0.5766 to a ten-seed ensemble of 0.6470, and three train-only post-hoc re-scorings targeting the metric bottleneck reach 0.6544. In contrast, an outlier-exposure component, architecture ablations, and from-scratch fusion models do not survive seed variance and are reported as negative results. The pipeline transfers unchanged to the all-real evaluation machines.

System characteristics
Classifier AE
System complexity 28177120, 56354240
Acoustic features log-mel energies
Decision making average
Subsystem count 10, 20
Front end system noise-aware far-mic reference denoising; cross front-end (per-band adaptive spectral subtraction [specsubada] and fixed-coefficient spectral subtraction [refsub]), dev-Omega-selected; per-domain-centered reconstruction-energy term, noise-aware far-mic reference denoising; fixed-coefficient spectral subtraction (RefSub, alpha=1.5 beta=0.10); per-domain-centered reconstruction-energy term, noise-aware far-mic reference denoising; per-band adaptive spectral subtraction (AdaSub); per-domain-centered reconstruction-energy term, noise-aware far-mic reference denoising; per-clip blend of AdaSub and inter-channel coherence-gate (CohGate); per-domain-centered reconstruction-energy term
PDF

Residual View and Prototype Selection for Noise-Aware Anomalous Sound Detection

JeongSik Kim, JongWoo Sung, HyeonJun Bae, BoRyeon Kim, JiAn Lee LUDO
Fundamental Deep Learning Research Group, LUDO Lab cooperation, Busan, South Korea and LUDO Lab cooperation, Busan, South Korea

Abstract

In this paper we take an in-depth look at noise-aware unsupervised anomalous sound detection in a GenRep-style frozen embedding memory-bank framework using pretrained audio encoders. We propose Residual View, which subtracts a scaled far-channel em-bedding from the near-channel embedding. Additionally, we use a projection-residual prototype selection branch for the submitted systems. Furthermore, we analyze the effect of the residual coef-ficient and representation layer. We also benchmark the proposed view with several pretrained audio encoders. Our final submitted systems apply PRPS to Residual View and achieve 63.24 official score with SSLAM on the development set.

System characteristics
Classifier kNN, pre-trained models
System complexity 21003776, 265732336, 89972736
Acoustic features log-mel energies
Data augmentation MemMixup
Decision making average
System embeddings AudioMAE++ tiny, BEATs iter3, DaSheng-base, SSLAM, SSLAM
Subsystem count 3
External data usage pre-trained model
PDF

ONLINE AND OFFLINE ENSEMBLE STRATEGIES FOR NOISE-A W ARE UNSUPERVISED ANOMALOUS SOUND DETECTION

Carl-Emil F . Krag, Christian Rhod, Emil Rosenlund, Simon Bøgh Bræck, Aalborg
Computer Engeneering, Aalborg University, Aalborg, Denmark

Abstract

This technical report details an ensemble pipeline submitted for the DCASE 2026 Challenge Task 2: Noise-Aware Unsupervised Anomalous Sound Detection for Machine Condition Monitoring. Our approach fine-tunes BEATs and SSLAM with an auxiliary classification objective where each combination of machine type, domain, attributes, and spatial audio channel is treated as a distinct class. This encourages the models to capture machine-specific acoustic characteristics while exploiting channel-dependent spatial information. For anomaly detection, the extracted embeddings are processed by multiple anomaly scorers, including KNN, GMMs, Autoencoders, and local density normalized KNN. The outputs from these individual detectors across the fine-tuned BEATs and SSLAM models are aggregated using a min-pooling score-based fusion strategy. To address both offline and real-time inference requirements, we employed rank normalization to stabilize scores in the offline setting, and calibrated Z-score normalization to maintain test sample independence in the online setting. The proposed ensemble substantially outperforms the baseline system under both offline (+10.69) and online (+8.67) evaluation protocols.

System characteristics
Classifier AE, GMM, kNN
System complexity 180284528, 270596323
Acoustic features log-mel energies, spectrogram
Data augmentation gaussian white noise, time shifting
Decision making minimum
System embeddings BEATs, SSLAM
Subsystem count 2, 3
PDF

DOMAIN-BLIND, DUAL-MICROPHONE ENSEMBLES FOR FIRST-SHOT ANOMALOUS SOUND DETECTION

Junhyeong Kwon, Jongsuk Choi, Korea
Korea Institute of Science and Technology, Seoul, Korea

Abstract

We describe our submission to DCASE 2026 Task 2 (Noise-aware Unsupervised Anomalous Sound Detection for Machine Condition Monitoring), the first edition to release two-channel near- and farmicrophone recordings, in the first-shot setting under domain shift. With only about a thousand normal clips per machine and no anomalies, we keep the analytical backbone frozen: a self-supervised BEATs encoder used as a fixed feature extractor, avoiding the representation collapse that full fine-tuning invites at this data scale. The core anomaly score is a per-domain-centred Mahalanobis ensemble over its embeddings. We exploit the second microphone in two unsupervised ways: a magnitude spectral-subtraction front-end that suppresses shared ambient noise, and a cross-channel difference stream from a second frozen encoder, Dasheng. Because each test clip’s domain is hidden at scoring time, we score every clip against both the source and target normal models and take the minimum- a domain-blind rule that is a correctness condition on the tagless evaluation set. Finally, as a separate decorrelated member we add a parameter-efficient LoRA–ArcFace encoder scored by cosine kNN, our largest single learned gain. On the development set the full ensemble reaches an official Ω of 68.97 %, a +19 % relative gain over the official Mahalanobis baseline; even our training-free core reaches 67.85 %. We submit four systems spanning a training-free baseline through the full learned ensemble.

System characteristics
Classifier ArcFace, LoRA fine-tuning, Mahalanobis distance, ensemble, kNN
System complexity 175801840, 90354032, 92948848
Acoustic features log-mel energies
Data augmentation mixup
Decision making weighted average
System embeddings BEATs, BEATs, Dasheng, BEATs, fine-tuned BEATs (LoRA)
Subsystem count 2, 3
External data usage pre-trained model
Front end system dual-microphone magnitude spectral subtraction
PDF

TASK-ADAPTED DUAL-MICROPHONE REPRESENTATIONS WITH DOMAIN-CONDITIONED LOCAL-DENSITY FUSION

Jialong Lei, CRRC, CRRC QINGDAO SIFANG CO
CRRC QINGDAO SIFANG CO., LTD., CRRC, Shandong, China

Abstract

Dual-microphone first-shot unsupervised anomalous sound detection requires a detector to separate weak machine faults from operating-condition changes and environmental interference, while only normal sounds are available for training. We build a multirepresentation detector around a parameter-efficient, task-adapted dual-microphone EAT encoder. Synchronized near and far channels share the pretrained backbone; low-rank attention updates and metadata-supervised objectives reshape intermediate statistics while preserving the transferable acoustic prior. At inference, the adapted statistics are scored against source- and target-domain normal memories using domain-conditioned local-density normalization and fused with reconstruction, CED, and M2D evidence. Controlled experiments show that covariance conditioning and task adaptation provide complementary gains: PCA whitening improves frozen EAT from 0.5834 to 0.6074, while DualMic-EAT reaches 0.6141±0.0004over two seeds (sample standard deviation 0.0004). Fixed-configuration outer leave-one-machine-out (LOMO) validation gives 0.6164, and the submitted four-branch cross-fitted fusion reaches 0.6250. A leave-one-branch-out study identifies M2D as the strongest complementary branch and motivates pruning the redundant frozen EAT anchor. The complete-development score of 0.6355 is reported only as a descriptive system-selection result.

System characteristics
Classifier AE, ArcFace, PCA whitening, attention LoRA, audio transformers, ensemble, kNN, local-density kNN, selective Mahalanobis distance
System complexity 300696, 353400000, 85793690
Acoustic features M2D-AS embeddings, log-mel energies, log-mel energies and LoRA-adapted transformer hidden states
Decision making weighted anomaly-score fusion (per-machine normal-reference quantile threshold), weighted average
System embeddings CED-Base, DualMic-EAT, M2D-AS, M2D-AS
Subsystem count 2, 4
External data usage pre-trained embedding models and parameter-efficient fine-tuning, pre-trained model, embeddings
Front end system least-squares far-channel noise residual, near microphone channel, synchronized near and far microphone channels
PDF

NOISE-AWARE UNSUPERVISED ANOMALOUS SOUND DETECTION WITH UBM AND GLOBAL GMM

Xiao Mei, Fund Intelligence Detection Co
Fund Intelligence Detection (Shanghai) Co., Ltd, Shanghai, China

Abstract

This report describes our submission to DCASE 2026 Task 2: Noise-aware Unsupervised Anomalous Sound Detection for Ma - chine Condition Monitoring. We propose three handcrafted sys - tems built on a 221-dimensional clip-level feature vector and Gaussian Mixture Models. A two-channel scaled-difference sig - nal suppresses environmental noise. Two systems use a Universal Background Model likelihood-ratio framework; the third uses a per-machine GMM. No evaluation labels or public code are used.

System characteristics
Classifier GMM, UBM likelihood ratio
System complexity 266950, 405144
Acoustic features handcrafted features, log-mel energies, nonlinear features
Decision making likelihood ratio, negative log-likelihood
Subsystem count 2
PDF

ML-SCAD: MULTI-LEVEL STEREO CONDITIONAL ANOMALY DETECTION FOR DCASE 2026 TASK 2

Junghyun Moon
Independent Researcher, Seoul, Republic of Korea

Abstract

This technical report describes ML-SCAD (Multi-Level Stereo Conditional Anomaly Detection), a system submitted to DCASE 2026 Task 2 that treats the synchronized two-channel stereo recording as a core design element rather than an auxiliary input. The near microphone (machine-dominant) provides the primary anomaly representation; the far microphone (noisedominant) is exploited in five complementary roles across signal, structural, semantic, score, and spectral levels. The system combines a BEATs/CLAP/autoencoder base ensemble with stereoaware modules: Wiener-filtered AE (L1), inter-channel relation LOF (L2), machine-specific BEATs routing with Far-channel Attribute-Conditioned Anomaly detection (FACA) and NoiseAware Score Normalization (NASN) (L3/L4), and Mel-Band Cross-Spectrum (MBCS) plus Cross-Channel Predictive Coding (CCPC) auxiliary branches (L5). MBCS approximates a stereo acoustic-transfer signature via ILD temporal variability (ILDσ) and magnitude-squared coherence (MSC), achieving standalone avg Ω=0.5950, outperforming BEATs-only Mahalanobis (Ω=0.5502) on all 7 development machines with identical hyperparameters. The submitted system (s final=0.85˜sbase+0.10˜sCCPC+0.05˜sMBCS) achieves per-machine avgΩ=0.6581(AUC src=0.7793, AUCtgt=0.6501, pAUC=0.6051; official single-HMΩ=0.6431) on the 7-machine development set. No anomaly labels are used for training, and evaluation labels are never accessed.

System characteristics
Classifier AE, KMeans, LOF, MLP, Mahalanobis distance, PatchCore, weighted ensemble
System complexity 248623282
Acoustic features GCC-PHAT, MFCC, energy ratio, inter-channel coherence, log-mel energies, mel-band cross-spectrum, spectral kurtosis
Data augmentation none
Decision making thresholding of the final anomaly score using a fixed threshold determined before evaluation
System embeddings BEATs, LAION-CLAP
Subsystem count 5
External data usage pretrained BEATs and LAION-CLAP models listed in the official allowed external resources; no external training dataset directly used
Front end system Wiener filtering, spectral subtraction, PCA, median-MAD score normalization
PDF

NOISE-A W ARE FIRST-SHOT UNSUPERVISED ANOMALOUS SOUND DETECTION VIA FROZEN ENCODERS, WIENER DENOISING, AND SHIFTALL-LDKNN SCORING

Yasaman Shokriazar, Arian Moradi, Felix Leber, Alfiia Ziganshina
Johannes Kepler University Linz, Linz, Austria

Abstract

We describe a system for DCASE 2026 Challenge Task 2: first-shot noise-aware unsupervised anomalous sound detection for machine condition monitoring. The system uses two frozen pre-trained audio encoders-BEATs and EAT-base-without any fine-tuning or gradient updates. Two-channel audio is processed via stereo Wiener denoising (using the far microphone as a noise reference) and a raw near – far difference view to suppress environmental noise. For anomaly scoring, we use local-density k-nearest-neighbour (LDKNN) heads. To overcome the extreme scarcity of target-domain training data, we apply a mean-shift augmentation (shiftall) that projects source-domain embeddings into the target domain, effectively expanding the small ten-sample target memory bank. The final anomaly score is a linear fusion of four complementary components using globally fixed weights, with no per-machine tuning or evaluation-score normalization. To ensure generalizability to unseen machines, validation follows a strict leave-one-machine-type-out (LOMO) cross-validation protocol. We submit two weight configurations. Our primary submission utilizesLOMO-consensusweights, obtained by averaging the optimal weights across all LOMO folds, ensuring weights are selected without ever seeing the full held-out fold scores (dev Ω = 0.6235). A secondary backup submission uses weights optimized directly on the full development set (Ω = 0.6256). The strict LOMO out-of-fold average harmonic mean is 0.6147 (worst fold0.5604).

System characteristics
Classifier PCA-Mahalanobis, ensemble, kNN, local-density KNN
System complexity 277000000
Acoustic features frozen BEATs embeddings, frozen EAT-base embeddings
Data augmentation shiftall embedding mean-shift augmentation
Decision making threshold at zero, weighted average
System embeddings BEATs, EAT-base
Subsystem count 4
External data usage pre-trained model embeddings
Front end system stereo Wiener denoising, near-far channel difference
PDF

MEMORY BANK-BASED UNSUPERVISED ANOMALOUS SOUND DETECTION EXPLOITING STEREO SPA TIAL INFORMA TION

Ryo Morita, Ryoya Kozai, Shota Sekino, Yusuke Kishi, KONICA MINOLTA
KONICA MINOLTA, INC., Tokyo, Japan

Abstract

This report describes our four systems submitted to DCASE 2026 Challenge Task 2: Noise-Aware Unsupervised Anomalous Sound Detection for Machine Condition Monitoring. All systems share a unified hypothesis: stereo spatial information is the key to domainrobust anomaly detection. Within a memory bank-basedk-nearest neighbor (kNN) framework, we present two contrasting strategies for exploiting the two-channel recording. Systems 1–2 employ a coherence-weighted power ratio mask for signal-level noise suppression combined with a fully fine-tuned EAT-large transformer trained with ArcFace loss. Systems 3–4 adopt a training-free approach using domain invariant features (DIF) extracted from stereo signal processing combined with frozen EAT-large embeddings. Our best system (System 1) achievesΩ = 64.41%on the development set, while the training-free System 3 achievesΩ = 62.34% with zero trainable parameters. The two designs-learned adaptation and physics-based invariance-provide complementary strategies that may generalize differently to unseen evaluation machines.

System characteristics
Classifier ZCA whitening, kNN
System complexity 0, 308000000, 311000000, 311000000 * 4
Acoustic features MFCC, band energy ratio, impulsiveness, log-mel energies, spectral contrast, stereo coherence profile
Data augmentation Temporal crop augmentation, mixup
Decision making average, weighted average
System embeddings EAT
Subsystem count 2, 4
External data usage Pre-trained model
Front end system Coherence and power-ratio mask
PDF

ADAPTIVE MULTI-PARADIGM ENSEMBLE FOR NOISE-A W ARE UNSUPERVISED ANOMALOUS SOUND DETECTION

Yu-jin Choi, Ji-sang Yoo, Dong-ha Oh, Seong-min Lee, Jin-yang Lee, Hee-seok Jeon, Jung-hoon Noh
School of Semiconductor Engineering, Chungbuk National University, Cheongju, Republic of Korea

Abstract

We present an adaptive multi-paradigm ensemble for noiseaware unsupervised anomalous sound detection. Our system combines eight orthogonal scorers-autoencoder (AE) reconstruction with Mahalanobis distance, BEATs-frozen k-NN, BEATs+LoRA discriminative k-NN, PaDiM-on-BEATs patchwise density, Modulation-Spectrum Mahalanobis, Cross-channel Coherence, and two Masked-Spectrogram-Modeling (MSM) variants (random pixel mask and curriculum frame-block mask)-each capturing a distinct aspect of normal-machine-sound semantics. A novel adaptive λ rule maps training-data-only signal features (Hilbert-envelope kurtosis, Ch0-Ch1 Pearson correlation, attributeCSV diversity) to per-machine ensemble weights, eliminating reliance on test-set labels at deployment time. On the DCASE 2026 Task 2 development set the submitted system achieves an official-score h-mean of 0.6300 (vs. 0.5789 baseline, +0.051 absolute). All score normalization uses domain-wise rank to neutralize source/target scale gaps.

System characteristics
Classifier AE, Mahalanobis distance, PaDiM, ensemble, kNN, masked spectrogram modeling, sub-cluster AdaCos
System complexity 3557832
Acoustic features cross-channel coherence, log-mel energies, modulation spectrum
Decision making weighted average
System embeddings BEATs
Subsystem count 8
External data usage pre-trained model
PDF

ANOMALOUS SOUND DETECTION METHOD WITH SIMPLE NOISE REDUCTION

Kosei Ozeki, Takeru Shiraga, Hideaki Terashima, Nobuaki Tanaka, and Takahiko Masuzaki, Mitsubishi Electric
Artificial Intelligence R&D Dept., Mitsubishi Electric Corporation, Kanagawa, Japan and Mitsubishi Electric Corporation, Kanagawa, Japan

Abstract

This paper presents anomalous sound detection methods for DCASE2026 Task 2. The goal of this contest is to identify whether the sounds emitted from target machines are normal or anomaly. First, we applied a simple noise reduction method. This method uses the far -channel as a noise reference signal and cancels the noise component mixed into the near-channel using a linear filter. Next, we applied the following four anomaly detection methods.

System characteristics
Classifier AE, kNN
System complexity 267928, 85253504, 90307808, 90324962
Acoustic features Mel spectrogram, log-mel filterbank
Decision making domain
System embeddings BEATs, CED
Subsystem count average
External data usage pre-trained model, embeddings
Front end system Linear filter, Normalization, Patchification, Linear filter, Normalization, log-Mel filterbank extraction
PDF

ANOMALOUS SOUND DETECTION SYSTEM FOR DCASE 2026 TASK 2 USING DUAL-CHANNEL SPECTRAL SUBTRACTION AND EFFICIENT AUDIO TRANSFORMER

Fan Chu, Mengui Qian
National Intelligent Voice Innovation Center, Hefei, China

Abstract

This report outlines our approach to noise-aware first-shot unsupervised anomalous sound detection for machine condition monitoring, developed for DCASE 2026 Task 2. Given the constraint of only having normal operational data, alongside the complexities of variable audio durations and the availability of two-channel recordings captured at different distances, our method focuses on leveraging dual-channel signal enhancement and a pre-trained Efficient Audio Transformer (EAT) for robust anomaly detection. Key components of our approach include applying spectral subtraction using the distant microphone as a noise reference for effective denoising, standardizing heterogeneous audio lengths to 16 seconds via audio looping with cross-fading to suppress padding artifacts, and extracting acoustic features via an EAT backbone finetuned with classification objectives to enhance generalization to unknown domains and complex acoustic environments. Anomalies are detected using a K-Nearest Neighbors (KNN)-based method by measuring the distance between each test sample embedding and its nearest neighbors in the training set; greater distances imply higher anomaly likelihood. Our approach achieved notable performance on the development set, demonstrating its effectiveness. The harmonic mean of the AUC for the target domain was 69.52% and for the source domain was 70.61%. Additionally, the harmonic mean of the Partial AUC values (p=0.1) was 56.84%. These results underscore the robustness and applicability of our methodology in detecting anomalous sounds in various operational contexts.

System characteristics
Classifier kNN
System complexity 87M
Acoustic features log-mel energies
Data augmentation mixup
Front end system Spectral Subtraction
PDF

AUDIOCC SYSTEM FOR DCASE 2026 TASK 2: FINE-TUNED AUDIO FOUNDA TION MODELS FOR NOISE-A W ARE MACHINE SOUND ANOMALY DETECTION

Xinhu Zheng, Junjie Li, Anbai Jiang, Wenrui Liang, Tianyu Liu, Shuwei Zhang, Yanmin Qian, Xie Chen, Cheng Lu, Pingyi Fan, Wei-Qiang Zhang, Jia Liu
Auditory Cognition and Computational Acoustics Lab, Shanghai Jiao Tong University, Shanghai, China and Department of Electronic Engineering, Tsinghua University, Beijing, China and Schools of Economy, North China Electric Power University, Beijing, China

Abstract

This report presents the AudioCC submission to DCASE 2026 Challenge Task 2 on noise-aware machine sound anomaly detection. Our submission studies how self-supervised audio foundation models can be adapted to machine-condition monitoring through task-oriented fine-tuning, angular-margin representation learning, and distance-based anomaly scoring. We submit four systems consisting of two single-model systems and two score-fusion systems. The submitted systems achieve a best harmonic mean of 65.33% on the development dataset.

System characteristics
Classifier ensemble, pre-trained models
System complexity 168000000, 22000000, 450000000, 90000000
Acoustic features log-mel filterbank
Data augmentation SpecAugment
Decision making Bayesian optimization
Subsystem count 3, 7
External data usage BEATs, FISHER-small
PDF

VUI LABS SYSTEM FOR DCASE 2026 TASK 2: DUAL-CHANNEL BEATS FUSION FOR MACHINE SOUND ANOMALY DETECTION

Yanmin Qian, Xinhu Zheng, Anbai Jiang, Wenrui Liang, Tianyu Liu, Shuwei Zhang, Xie Chen, Cheng Lu, Wei-Qiang Zhang, Pingyi Fan, Jia Liu
VUI Labs, Shanghai, China and Shanghai Jiao Tong University, Shanghai, China and Tsinghua University, Beijing, China and North China Electric Power University, Beijing, China

Abstract

This report describes the VUI Labs submission to DCASE 2026 Challenge Task 2 on noise-aware anomalous sound detection. The submission focuses on dual-channel modeling for paired near-field and far-field machine recordings. We combine single-channel self-supervised representations with channel-interaction variants that use reference-channel information to stabilize noisy-channel anomaly scoring. Four systems are submitted, including two singlemodel systems and two score-level fusion systems. The best submitted system achieves a harmonic mean of 66.39% on the development dataset.

System characteristics
Classifier ensemble, pre-trained models
System complexity 134000000, 180000000, 90000000
Acoustic features log-mel filterbank
Data augmentation SpecAugment
Decision making Bayesian optimization
Subsystem count 2, 5
External data usage BEATs
PDF

A SPATIAL-TEMPORAL ATTENTION AND CONFIDENCE-BASED DOMAIN ADAPTATION FRAMEWORK FOR NOISE-A W ARE FIRST-SHOT UNSUPERVISED ANOMALOUS SOUND DETECTION

Nighil Natarajan, Raghav Sridharan, Nithilan M, Chandrakala S
Department of Computer Science and Engineering, Shiv Nadar University Chennai, Chennai, India

Abstract

This report presents our system for the DCASE 2026 Challenge Task 2 on noise-aware first-shot unsupervised anomalous sound detection. We propose a domain-adversarial contrastive learning framework that learns robust acoustic representations under domain shift and varying machine operating conditions. Raw audio captured from the near microphone is converted into RGB log-Mel spectrogram patches and processed by a pre-trained ResNet-34 encoder. Spatial features are extracted through attention pooling, while a temporal encoder captures sequential acoustic dependencies. To improve robustness against domain shifts and environmental noise, the framework combines confidence-weighted contrastive learning with Domain Adversarial Neural Networks (DANN), CORAL, and Maximum Mean Discrepancy (MMD) losses. Anomaly detection is performed using a hybrid Mahalanobis-Cosine scoring strategy over clustered normal embeddings. Experimental results on the DCASE 2026 Task 2 development dataset demonstrate improved anomaly detection performance under noise-aware and domain-shift conditions. The source code is publicly available athttps: //github.com/nighiln05/DCASE_26_Task2.

System characteristics
Classifier ResNet34, attention pooling, contrastive learning, domain adversarial neural network (DANN), temporal convolutional network
System complexity 22000000
Acoustic features RGB log-Mel spectrogram
Data augmentation random resized crop, random horizontal flip, color jitter, random grayscale
System embeddings ResNet34 ImageNet pretrained embedding, projection head embedding, fused spatial-temporal embedding
Subsystem count 1
Front end system log-Mel spectrogram, RGB colormap conversion, image resizing
PDF

FOUR SYSTEMS FOR NOISE-A W ARE UNSUPERVISED ANOMALOUS SOUND DETECTION IN DCASE 2026 TASK 2

Tsz Fhl
TODO, HFU, HEFEI, China

Abstract

This technical report describes four submissions to DCASE 2026 Task 2 [1]: Tsz DCASE task2 baseline (official autoencoder baseline with Mahalanobis scoring), Tsz DCASE task2 1 (stereo discriminative GeoSNet with BEATs and prototype scoring), Tsz DCASE task2 fusion (fine-tuned BEATs with ArcFace and rank-min fusion), and Tsz DCASE task2 sera (SERA-style ASP with frozen BEATs and multi-task ArcFace). All systems operate on 16 kHz stereo recordings under domain shift [2, 3, 4], use per-machine or global training on development and eval-additional data, and apply gamma-fit decision thresholds (q=0.9) on trainnormal scores. Development-set results are reported for architecture validation; eval submissions provide anomaly scores and binary decisions for five official machine types.

System characteristics
Classifier ASP + ArcFace multi-task (machine + attribute heads), ArcFace multi-class training on composite labels; rank-normalized KNN+GMM fusion scoring, CNN+MSFE, DCASE2023T2 autoencoder, FiLM+Conformer, SGFF fusion, discriminative embedding, per-machine training on normal clips, prototype min-distance scoring
System complexity 1000000, 70000000, 95000000
Acoustic features BEATs encoder patch sequence, generalized mean pooling, log-mel energies, spatial cues
Data augmentation gaussian noise, time shift, mixup, target oversampling
Decision making gamma-fit threshold on train-normal fusion scores (q=0.9), gamma-fit threshold on train-normal proto scores (q=0.9), gamma-fit threshold on train-normal scores (q=0.9)
System embeddings 768-d clip embedding after top-4-layer BEATs fine-tune, Attentive Statistics Pooling + BN, BEATs (frozen, near-mic waveform)
Subsystem count 2
External data usage BEATs iter3+AS2M (AudioSet) pretrained checkpoint, BEATs iter3+AS2M finetuned checkpoint (AudioSet), pre-trained BEATs checkpoint (AudioSet)
Front end system frozen BEATs (near-mic mono waveform), frozen/finetuned BEATs (near-mic mono, 10s clip), log-mel spectrogram (128 mels, 5-frame stack), stereo log-mel + spatial cues (ILD/ITD/IPD/MS), PCEN optional path
PDF

NOISE-AWARE UNSUPERVISED MACHINE ANOMALOUS SOUND DETECTION USING PRE-TRAINED ACOUSTIC REPRESENTATIONS, DUAL-CHANNEL GATED FUSION, AND NORMAL-REFERENCE MODELING

Lianzhi Wang, Jeffrey Liu, Suzhou Dongyuan Electronics Co
Suzhou Dongyuan Electronics Co., Ltd., China

Abstract

This report describes our system for DCASE 2026 Challenge Task 2, noise-aware unsupervised anomalous sound detection for machine condition monitoring. The task requires a system to use only normal training clips and to output continuous anomaly scores for unlabeled clips in the evaluation dataset; auxiliary decision files are generated only to match the submission package format. The system must address source-target domain shift, two-channel noisy recordings, a very small number of target-domain normal clips, and the generalization risk caused by the mismatch between development and evaluation machine types. We use pre-trained acoustic representation models as audio encoders and adapt them with Low-Rank Adaptation (LoRA). The main branch is based on BEATs, while the auxiliary branch is based on AudioMAE. Both branches use a dual-channel gated fusion module to map synchronized near-field and far-field channels into sample-level embeddings. During training, ArcFace-style attribute and domain classification objectives are used to shape the embedding space of normal samples. During inference, the classification heads are removed, normal training samples are used to build reference banks, and anomaly scores are computed by distance-based normal-reference scoring. Because the raw anomaly scores produced by different scoring backends have different scales, we calibrate the scores from the two branches by robust z-score calibration estimated only from normal reference scores, and then perform score-level fusion. On the development dataset, the BEATs-AudioMAE robust-z weighted ensemble with BEATs weight 0.85 and left-channelduplicated inference obtains an overall harmonic mean of 61.43%. This result is higher than the official simple autoencoder baseline of 56.66% and the selective Mahalanobis baseline of 57.66%. To reduce overfitting risk on the hidden evaluation set, we also submit a default two-channel inference variant, a more conservative fusion variant with a higher BEATs weight, and a BEATs single-encoder variant.

System characteristics
Classifier ArcFace representation learning, LoRA-adapted pre-trained BEATs encoder, LoRA-adapted pre-trained audio encoders, Mahalanobis distance, kNN cosine distance, score-level ensemble
System complexity 178000000, 91512832
Acoustic features raw waveform input to pre-trained BEATs and AudioMAE encoders, raw waveform input to pre-trained BEATs encoder
Data augmentation channel dropout, channel gain perturbation, channel swap
Decision making AudioMAE weight 0.10, AudioMAE weight 0.15, BEATs weight 0.85, BEATs weight 0.90, robust-z weighted average
System embeddings AudioMAE, BEATs, BEATs
Subsystem count 2
External data usage pre-trained model, pre-trained models
Front end system fixed-length waveform crop and padding, inference-time left-channel duplication, fixed-length waveform crop and padding, original dual-channel gated inference
PDF

UNIS SYSTEM FOR NOISE-A W ARE UNSUPERVISED ANOMALOUS SOUND DETECTION FOR MACHINE CONDITION MONITORING

Junjie Wang

Abstract

This technical report presents our solution to Task 2 of the DCASE 2026 Challenge. We developed four subsystems for unsupervised anomalous sound detection, all of which extract audio embeddings from fine-tuned audio pre-trained models and perform anomaly detection using outlier detection algorithms. Compared with previous editions of the challenge, the target machine sounds in this year’s task were synchronously recorded by multiple microphones, providing both near-field and far-field acoustic views. To exploit this characteristic, we leverage far-field recordings as a data augmentation strategy, enabling the models to learn more robust acoustic representations and improving their generalization ability under different recording conditions and noisy environments. In addition, we employ multiple pre-trained models to obtain complementary feature representations, further enhancing the overall anomaly detection performance.The source code is publicly available at:https: //github.com/outman-goutian/dcase2026_task2.

System characteristics
Classifier KNN, kNN
System complexity 180M, 90M
Acoustic features log-mel energies
Data augmentation mixup
System embeddings BEATs, BEATs, EAT, BEATs，EAT, EAT-base
Subsystem count 2
PDF

ENSEMBLE SYSTEM INCLUDING PRE-TRAINED MODEL BASED PROTOTYPES FOR DCASE2026 TASK2

Yanfei Wang Yongji Sun, FangPing Xie, Shanghai Wangshuo Technology Co
Shanghai Wangshuo Technology Co., Ltd, China

Abstract

In this report, we propose EAT based ensemble system to address the Dcase2026 Task 2. We used the pre-trained EAT model and fine-tuned it in the development set. Then, we built a prototype classifier and use the distance to prototypes to get the anomaly score. The system is well generalized and are easy to deploy. The final results are obtained through model ensemble by combining several models including the aforementioned ones, the official baseline and so on. Our final ensemble system has achieved 62.12% in the official scores calculated as a harmonic mean of the area under the curve (AUC) and partial AUC (p = 0.1) over all machine types and domains in the development set.

System characteristics
Classifier AE, EAT, GMM, LOF, baseline MAHALA, baseline MSE, ensemble, prototype, stereo AE
Acoustic features log-mel energies
Decision making percentile threshold
System embeddings EAT
External data usage pre-trained EAT model
PDF

UNSUPERVISED ANOMALOUS SOUND DETECTION VIA EA T AND KALMAN-FILTERED NEAR- AND FAR-FIELD DUAL-CHANNEL DA TA

Hao Wu, Zhansai Chang, Pengyuan zhao, Tianju Zhao, Yutao Zhang, Meng Lei, Liang Zou
China University of Mining and Technology, XuZhou,China

Abstract

This report presents our approach for DCASE 2026 Task 2 on firstshot unsupervised anomalous sound detection. To address complex acoustic environments, we propose a dual-channel fine-tuning framework utilizing an Efficient Audio Transformer (EAT). Our model is trained on both frequency-domain Kalman-filtered nearfield audio and original far-field audio. To handle varying durations, log-mel filterbank (fbank) features are zero-padded to ensure uniform dimensions. During feature extraction, representations are exclusively obtained from the filtered near-field channel and subjected to precise regularization: values below 0.2 are clamped to zero, while values above 0.5 are suppressed via a tanh function.Final anomaly scores are calculated using K-Nearest Neighbors (KNN). Our approach achieved notable performance on the development set, demonstrating its effectiveness. The AUC for the target domain was 71.33% and for the source domain was 72.28%. Additionally, the Partial AUC values (p=0.1) for the target and source domain were 58.20%. These results underscore the robustness and applicability of our methodology in detecting anomalous sounds in various operational contexts.

System characteristics
Classifier kNN
System complexity 87M
Acoustic features log-mel energies
Data augmentation mixup
Front end system Kalman filter
PDF

DISCRIMINATIVE RESNET AND BEATS ADAPTATION FOR NOISE-A W ARE MACHINE SOUND ANOMALY DETECTION

Jiakun Xia, Northeastern
Northeastern University, China

Abstract

DCASE2026 Task2 evaluates first-shot unsupervised anomalous sound detection under unseen machine types and noisy two-channel recordings. We submit four systems that share a common training, embedding, and scoring pipeline: a trainable MultiResNet spectral encoder, two adapted BEATs encoders, and a BEATs LoRA adaptation system. The first system can use channel 2 as a deterministic far-field noise reference during inference preprocessing, while the three BEATs systems are submitted with the raw channel-1 inference view. All systems use sub-cluster AdaCos supervision with 16 subclusters and cosine nearest-neighbor scoring against normal training embeddings.

System characteristics
Classifier BEATs embedding extractor with rank-128 LoRA modules, and cosine k-nearest-neighbor anomaly scoring, cosine k-nearest-neighbor anomaly scoring, sub-cluster AdaCos classifier
System complexity 5410033, 94389761, 96749057
Acoustic features FFT magnitude spectrum and STFT magnitude spectra, pre-trained audio embeddings
Data augmentation mixup
Decision making a fixed 0.9 normal-train-score quantile threshold, continuous cosine nearest-neighbor anomaly score (0.9 normal-train quantile threshold), fixed 0.9 normal-train-score quantile threshold
System embeddings BEATs
Subsystem count 3
External data usage BEATs pre-trained audio representation model
Front end system Raw wav ch1, Raw wav ch1, Wiener ch2
PDF

AN ENHANCED TRAINING-FREE ASD METHOD FOR NOISE-AWARE UNSUPERVISED ANOMALOUS SOUND DETECTION FOR MACHINE CONDITION MONITORING

Qin Xie, Luowei Ma, Shiyang Pei, Qinghua Huang
ShangHai University, ShangHai, China

Abstract

This report describes our method for DCASE 2026 Task 2: Noiseaware Unsupervised Anomalous Sound Detection (ASD) for Machine Condition Monitoring. Our method follows a training-free framework based on frozen audio embeddings and memory-bankbased anomaly scoring. To better utilize the noise -aware dual - channel recordings provided in the challenge, we introduce a difference-based feature enhancement strategy tha t combines near - field and far-field audio information. Furthermore, we investigate several temporal pooling methods to improve the aggregation of frame-level embeddings extracted by a frozen Efficient Audio Transformer (EAT). Experimental results show that both the proposed feature enhancement and temporal pooling strategies provide consistent improvements over the baseline system. The final system achieves enhanced anomaly detection performance while preserving the simplicity of training-free ASD methods.

System characteristics
Classifier kNN
System complexity 90,379,535
Acoustic features log-mel energies
Decision making minimum
System embeddings EAT
PDF

CHANNEL-ROBUST EMBEDDING ENSEMBLES FOR DCASE 2026 TASK 2

Xing Wu
not applicable, MCPX, not specified

Abstract

DCASE 2026 Task 2 evaluates first-shot unsupervised anomalous sound detection for machine condition monitoring under unseen machine types and noisy two-channel recordings. This report describes four submitted systems built from three complementary normal-sound embedding models: a tiny-size FISHER industrial-signal adapter, an EAT audiotransformer adapter trained with an angular-margin objective, and a small-size FISHER industrial-signal adapter. The primary system uses only the near microphone and fuses the three model scores after fixed machine-wise normalization. The remaining systems add a conservative Wiener-enhanced channel, a single-model near-microphone fallback, and an aggressive channel-2 hedge. All systems use normal-only embedding learning and cosine one-nearest-neighbor scoring against source and target normal references; larger scores indicate sounds farther from the normal training manifold.

System characteristics
Classifier EAT ArcFace embedding with cosine KNN, EAT/FISHER embeddings with cosine KNN
System complexity 119322368, 92540672
Acoustic features audio embeddings
Data augmentation mixup
Decision making normal-training score threshold with an expected 5% abnormal rate on normal training clips
System embeddings EAT, EAT and FISHER models
Subsystem count 3, 6, 9
External data usage EAT and FISHER pretrained models, EAT pretrained model
Front end system Ch1 waveform; optional Ch2-guided preprocessing
PDF

DISTILLATION GUIDED TWO-CHANNEL BEATS SYSTEM FOR DCASE2026 TASK 2

Juncai Yang, Nanjing
Nanjing University, Nanjing, China

Abstract

This technical report describes our submitted system for DCASE2026 Challenge Task2, Noise-aware Unsupervised Anomalous Sound Detection for Machine Condition Monitoring. The system is based on a pre-trained BEATs representation and a pseudoclean distillation strategy for two-channel machine-sound recordings. In the first stage, ILRMA-based blind source separation is applied to the two-channel recordings to obtain pseudo-clean waveforms, which are used to adapt a BEATs teacher model. In the second stage, the fixed teacher supervises a student model trained on the original two-channel recordings. The student model consists of a neural enhancement front-end, a shared BEATs encoder, an enhanced-mixture sequence fusion module, and attentive statistics pooling. Utterance-level embeddings extracted by the student model are finally scored by a KNN backend with cosine distance. The submitted system achieves an official score of 62.27% on the development set.

System characteristics
Classifier BEATs, LoRA fine-tuning, attentive statistics pooling, kNN, neural enhancement, teacher-student knowledge distillation, two-channel enhanced-noisy fusion
System complexity 101113203
Acoustic features log-mel energies, raw waveform
Data augmentation SpecAugment
System embeddings BEATs
External data usage pre-trained model
Front end system neural enhancement front-end
PDF

DUAL-CHANNEL A TTRIBUTE FINE-TUNING OF PRE-TRAINED MODELS FOR NOISE-A W ARE ANOMALOUS SOUND DETECTION

Jie Yang
Individual., HuaiBei, China

Abstract

This report details our technical approach to the DCASE 2026 Challenge Task 2, which focuses on Noise-aware Unsupervised Anomalous Sound Detection (UASD) for machine condition monitoring. In real-world scenarios, obtaining clean recordings of target machines or isolated background noise is often difficult because machines cannot be easily stopped. To provide a practical alternative for developing noise-robust systems under such constraints, the 2026 dataset introduces synchronous two-channel audio samples captured at locations near and far from the target machine, where the distant microphone serves as a noise reference. Our implemented system utilizes the Efficient Audio Transformer (EAT) base model, pre-trained on AudioSet-2M, as the robust feature extraction backbone. To exploit the dual-channel nature of the data, the two channels are explicitly separated to double the training data volume and zero-padded to a uniform length. The EAT model extracts deep features, followed by Attentive Statistics Pooling (ASP) for dimensionality reduction. We construct a unified composite label for each sample encompassing its machine type, intrinsic attribute, domain, and the designated ”near” or ”far” spatial tag. The backbone is then fine-tuned using the ArcFace loss function to maximize intra-class compactness. For anomaly scoring, a K-Nearest Neighbors (KNN) model is employed in the latent space. We submitted four systems combining different fine-tuning strategies and inference approaches. Evaluated on the development set, the proposed method yields a source domain AUC of 66.70%, a target domain AUC of 68.48%, and a pAUC of 59.05%.

System characteristics
Classifier kNN
System complexity 87M
Acoustic features log-mel energies
Data augmentation mixup
PDF

XJU SYSTEM FOR UNSUPERVISED ANOMALOUS SOUND DETECTION WITH PRE-TRAINED AUDIO MODELS AND DENOISING

Zhou Yang, XinJiang
School of Computer Science and Technology, XinJiang University., Urumqi, China and Tsinghua University, Beijing, China

Abstract

DCASE 2026 Task 2 focuses on noise-aware unsupervised anomalous sound detection for machine condition monitoring, where systems are required to detect unknown anomalies using only normal training data under noisy and domain-shift conditions. In this technical report, we present the XJU systems based on pre-trained audio model fine-tuning, domain-specific feature learning, denoisingbased front-end processing, and ensemble learning. The submitted systems include an OSSCL system, a BEATs-GRL system, an ANC-enhanced system, and an OSSCL-GRL ensemble system. On the development set, the submitted systems outperform the official baselines in terms of the official hmean score, and the ensemble system obtains an hmean score of 65.8%. In addition, we discuss training-time two-channel fusion as a possible future direction for noise-aware anomalous sound detection.

System characteristics
Classifier BEATs, EAT
System complexity 90496240, 907679660, 91039692
Acoustic features log-mel energies
Data augmentation mixup, SpecAugment
Decision making average
Subsystem count 10
Front end system ANC
PDF

A CONSTRAINT-DRIVEN STA TISTICAL PIPELINE FOR NOISE-A W ARE ANOMALOUS SOUND DETECTION UNDER FREE-TIER COMPUTE

Mehdi Zarrouky
Independent Researcher, Casablanca, Morocco

Abstract

We describe a fully statistical anomaly detection system submitted to DCASE 2026 Challenge Task 2 [3], developed under severe computational constraints (free-tier cloud notebooks, CPU only, no persistent compute environment). These constraints precluded the fine-tuning of large pre-trained acoustic models and shaped a methodology focused on parameter-free statistical inference rather than learned representations. The pipeline combines dual-channel exploitation (log-mel and constant-Q from both microphones and their difference), multi-scale temporal pooling, L2-normalized domain-aware Mahalanobis scoring with Ledoit-Wolf shrinkage [5], and rank-based ensemble. The system uses no neural networks and no external datasets. On development data, including ToyADMOS2 [1] and MIMII-derived machines [2], and following the first-shot anomaly detection paradigm [4], the system achieves a harmonic mean of 60.76 across AUC and pAUC atp= 0.1, approximately four points above the official Selective Mahalanobis baseline. Performance is strongest on stationary signals (fan, valve) and weakest on impulsive transients (gearbox, slider).

System characteristics
Classifier Mahalanobis distance with Ledoit-Wolf shrinkage
System complexity 20696340
Acoustic features CQT, log-mel energies
Decision making median threshold
Subsystem count 2
Front end system dual-channel (ch1, ch2, ch1-ch2)
PDF

Multi-System Framework with Heterogeneous Feature Extractors for DCASE 2026 Task 2

Wen Zeng, Jianxia Liao, Ting Wu, Zhaoli Yan, Fusheng Sui, 1Beijing
Beijing University of Chemical Technology, Beijing, China and Institute of Acoustics, Chinese Academy of Sciences, Beijing, China

Abstract

This report describes our submission to DCASE 2026 Challenge Task 2: First-Shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring. We propose a multi-system framework built upon heterogeneous feature extractors, including a BEATs -based audio encoder, a convolutional neural network (CNN) feature extractor adapted from the DCASE2023 FKIE system, and a ResNet-based audio encoder. Stereo information is exploited through both independent channel modeling and joint two -channel representation learning. Domain -wise Local Density Normalization (DLDN) is employed in the BEATs- and FKIEbased systems for anomaly score computation. Finally, anomaly scores from different channels and feature extractors are fused using optimized weights determined on the development dataset. Experimental results demonstrate that exploiting stereo -channel information and combining heterogeneous feature representations effectively improves robustness and detection performance under domain -shift conditions.

System characteristics
Classifier BEATs asp arcface, CNN
System complexity 2,188,544, 90000000
Acoustic features log-mel energies, log-mel filterbank
Data augmentation mixup, mixup, SpecAugment
Decision making average
System embeddings BEATs
PDF

MULTI-CHANNEL BRANCH CONSTRUCTION AND GRAPH-REFINED MEMORY BANKS FOR DCASE 2026 CHALLENGE TASK 2

Zhang Cheng, Masashi Unoki, Graduate
Japan Advanced Institute of Science and Technology, Nomi, Japan

Abstract

This technical report presents a method for DCASE 2026 Challenge Task 2, which focuses on unsupervised anomalous sound detection under noisy two-channel recording conditions. The key challenge is how to effectively exploit multi-channel information and improve memory-bank based anomaly scoring. We exploit the multi-channel nature of the dataset by constructing multiple channel embedding branches, including single-channel, mono, and inter-channel difference representations. To improve nearest-neighbor based anomaly scoring, we refine source and target memory banks with graph smoothing, align test samples to the refined memory structure at inference time, and apply local-density normalized KNN scoring. Finally, scores from multiple pretrained encoder branches are fused at the score level. On the DCASE 2026 development set, the proposed system achieves an official score of 0.661203, outperforming both the official baseline and the GenRepASD baseline. Experimental results show that multi-channel branch construction is the main contributor, while graph refinement, test-time graph alignment, and density-normalized KNN score provide additional consistent gains. Our source code is available athttps://github. com/infolence/DCASE2026_Task2.git.

System characteristics
Classifier kNN
System complexity 96220238
Acoustic features raw waveform
Decision making weighted average
System embeddings BEATs, CED
Subsystem count 4
External data usage pre-trained model
PDF

SA TLAB SYSTEM FOR DCASE 2026 TASK 2: GENERA TIVE DA TA AUGMENTA TION FOR MACHINE SOUND ANOMALY DETECTION

Wenrui Liang, Tianyu Liu, Xinhu Zheng, Anbai Jiang, Shuwei Zhang, Pingyi Fan, Cheng Lu, Yanmin Qian, Xie Chen, Jia Liu, Wei-Qiang Zhang, 1Tsinghua
Tsinghua University, Beijing, China and Shanghai Jiao Tong University, Shanghai, China and North China Electric Power University, Beijing, China

Abstract

This report describes the SATLab submission to DCASE 2026 Challenge Task 2 on noise-aware anomalous sound detection. The central component of our system is a generative data augmentation pipeline: we train a diffusion-based audio generator from scratch on the challenge data to synthesize samples of rare working conditions, and combine the screened synthetic samples with the real recordings to alleviate the data imbalance and the source–target domain gap. On top of this augmented training set, the submitted systems use BEATs-based audio representations, Mahalanobis and nearest-neighbor anomaly scoring, and score-level ensemble selection. Four systems are submitted, including one single scoring system and three ensemble systems. The best submitted system achieves a development-set harmonic mean of 69.43%.

System characteristics
Classifier diffusion, ensemble, pre-trained models
System complexity 1080000000, 630000000, 810000000, 90000000
Acoustic features log-mel filterbank
Data augmentation SpecAugment
Decision making weighted average (Bayesian optimization)
System embeddings BEATs
Subsystem count 12, 7, 9
External data usage BEATs (pre-trained model)
PDF

THUEE SYSTEM FOR DCASE 2026 TASK 2: ENSEMBLING MULTIPLE ANOMALY DETECTORS FOR GENERALIZED ANOMALOUS SOUND DETECTION

Shuwei Zhang, Wenrui Liang, Xinhu Zheng, Lvxin Xu, Anbai Jiang, Tianyu Liu, Pingyi Fan, Wei-Qiang Zhang, Yanmin Qian, Xie Chen, Cheng Lu, Jia Liu
Tsinghua University, Beijing, China and Shanghai Jiao Tong University, Shanghai, China and North China Electric Power University, Beijing, China

Abstract

This report presents the THUEE submission to the DCASE 2026 Challenge Task 2, which addresses noise-aware anomalous sound detection (ASD) for machine condition monitoring. The task introduces a dual-channel setup where one microphone captures near-field machine sound and the other captures far-field, noisecontaminated recordings. Our system builds on an audio foundation model, OpenBEATs, and fine-tunes it with a joint classification objective over machine type, operating attribute, and channel index. To enhance generalization, we augment the training data with pseudo spectrograms generated by a generative model. For anomaly detection, we extract utterance-level embeddings and employ five diverse scoring strategies: three KNN-based detectors (including local density rescaling and VarMin), a Mahalanobis-distance-based detector, and a flow-matching-based detector. Anomaly scores from these detectors are fused through weighted linear combination optimized via Bayesian search. Four system variants are submitted, among which the ensemble combining KNN, Mahalanobis, and flow matching detectors achieves the best overall harmonic mean of 69.54% on the development set. We further observe that although the far-field channel benefits representation learning during fine-tuning, it tends to introduce distraction during anomaly scoring, highlighting the need for careful channel-level treatment.

System characteristics
Classifier Mahalanobis distance, ensemble, kNN, pre-trained models
System complexity 1440000000, 2070000000, 630000000, 90000000
Acoustic features STFT, log-mel filterbank
Data augmentation SpecAugment
Decision making Bayesian optimization
Subsystem count 16, 23, 7
External data usage BEATs
PDF

ROBUST TRAIN-NORMAL CALIBRA TION FOR FIRST-SHOT MACHINE SOUND ANOMALY DETECTION

Peihong Zhang Shengchen Li, Xi’an Jiaotong-Liverpool
School of Advanced Technology, Xi'an Jiaotong-Liverpool University, Suzhou, China

Abstract

DCASE 2026 Task 2 evaluates first-shot unsupervised anomalous sound detection under source-domain, target-domain, and low-FPR criteria. In this setting, our experiments indicate that the dominant failure mode is not simply insufficient representation capacity, but unstable normal-score tails and fragile source/target operating conditions. We therefore submit a frozen normality-calibration portfolio: complementary pretrained audio scores are normalized only with train-normal references, guarded by conservative tail and domain-risk reliability checks, and evaluated with an exact replica of the official development metric. The primary system improves a strong proxy score from 0.611706 to 0.613941 official harmonic mean while preserving mean pAUC at 0.592030. Additional submitted slots provide an out-of-fold mirror, an independent Dasheng guarded variant, and a global-gating fallback for target false-positive risk. Negative stress tests are also informative: larger resource ensembles, adaptive local-density normalization, and localized temporal matching improved some AUC or target-risk indicators but degraded low-FPR pAUC, so they were excluded. The resulting submission emphasizes a practical conclusion for first-shot ASD: robust train-normal calibration and normal-tail preservation can be more valuable than adding another high-capacity detector.

System characteristics
Classifier reliability gating, score ensemble, train-normal density scoring
Acoustic features frozen audio embeddings, train-normal kNN distance scores
Decision making fixed train-normal score threshold
System embeddings CED-small, Dasheng, EAT
Subsystem count 3
External data usage pre-trained embeddings
Front end system near channel, mono mean channel
PDF

ABNORMAL SOUND DETECTION BASED ON NOISE-AWARE DOMAIN GENERALIZATION

Zhifang Zheng, Shengbing Chen, Yanjun Zhou, Younan Ji, Hanbin Zhou, Shuchi Chen
School of Artificial Intelligence and Big Data, Hefei University, Hefei, China

Abstract

This technical report describes our anomalous sound detection systems submitted for DCASE 2026 Challenge Task 2. The task focuses on noise-aware unsupervised anomalous sound detection for machine condition monitoring, where only normal machine sounds are available for training and the test data may include domain shifts and environmental noise. To address this setting, we build several simple systems based on pre-trained audio representations, two-channel noise-aware auxiliary features, and distance-based anomaly scoring. The submitted systems mainly differ in the combination of BEATs embeddings, EAT embeddings, and spatial features extracted from near- and farmicrophone recordings. For each machine type, normal training samples are used as reference data, and anomaly scores are obtained by measuring the distance between test samples and the reference normal samples. The final submitted systems are selected on the development set according to the official metrics.

System characteristics
Classifier kNN, score fusion
System complexity 182092927, 91713392
Acoustic features BEATs LoRA embeddings, EAT embeddings, spatial statistics
Data augmentation noise mixing, random gain, time shift
Decision making weighted average of normalized anomaly scores
System embeddings BEATs, BEATs, EAT
Subsystem count 2, 3, 4
External data usage pre-trained BEATs and EAT models, pre-trained BEATs model
Front end system near-far microphone spatial statistics
PDF

EAT TOKEN-MFS/GBC BEAM ANOMALOUS SOUND DETECTION SYSTEM FOR DCASE2026 TASK 2

Hanbin Zhou Shuchi Chen Shenbin Chen Zhifang Zheng Shuzheng Tang
HFUU_DSAI, Hefei, China

Abstract

This report describes our anomalous sound detection system for DCASE 2026 Task 2. The system uses a pretrained Efficient Audio Transformer (EAT) backbone to extract audio token representations from mel-spectrograms. A token-level multi-frame-scale branch with gated bottleneck convolution (MFS/GBC) is used to enhance local time-frequency structures, and LoRA is adopted for parameter-efficient adaptation. At inference time, normal training embeddings are stored in a beam-subband memory bank, and anomaly scores are computed with a nearest-neighbor backend. On the development set, the system obtains 69.28% mean AUC, 75.21% source AUC, 63.35% target AUC, 58.05% pAUC, and a 63.48% official score.

System characteristics
Classifier ArcFace, EAT, LoRA, kNN, memory bank
System complexity 304000000
Acoustic features EAT token maps, log-mel energies
Data augmentation supplemental mixup
Decision making thresholding on anomaly score
System embeddings EAT large pretrained on AudioSet-2M
External data usage pretrained model
Front end system Mel spectrogram
PDF

MACHINE-WISE DUAL-CHANNEL ANOMALOUS SOUND DETECTION WITH TARGET-A W ARE THRESHOLDING

Yunhong Zhou, SUMERU ZOO, Beijing, China
SUMERU ZOO, Beijing, China

Abstract

We present a machine-wise anomalous sound detection system for DCASE 2026 Task 2. The proposed system addresses the challenges of first-shot unsupervised detection under severe source-target domain shift and dual-channel recording conditions. Instead of applying a single unified strategy to all machines, we separate the pipeline into two branches: conditional routing for attribute-aware machines and global healthy-reference modeling for no-attribute machines. The submitted system adopts a per-machine sub-band representation, followed by Mahalanobis-based anomaly scoring against a healthy reference. For decision-result generation, we further introduce a target-aware thresholding rule that relies only on legal train-normal data and helps reduce domain-induced false positives on machines with strong target shift. In addition, diagnostic side experiments on development machines were used to understand error modes and support system design. The resulting framework provides an interpretable and robust machine-wise solution for firstshot anomalous sound detection.

System characteristics
Classifier Mahalanobis distance, cosine distance
System complexity 110000000, 5000
Acoustic features log-mel energies, spectral and temporal statistics
Data augmentation none
Decision making maximum
System embeddings BEATs
Subsystem count 4
External data usage pre-trained model
Front end system none
PDF

MACHINE ANOMALOUS SOUND DETECTION USING NOISE-A W ARE SPECTRAL SUBTRACTION AND CONFORMER BASED FEA TURE LEARNING

Qing Zhou, Shan Li, Xi’an
College of Information and Control Engineering, Xi'an University of Architecture and Technology, Xi'an, China and Xi'an University of Architecture and Technology, Xi'an, China

Abstract

DCASE2026 Task 2 focuses on unsupervised anomalous sound detection for machine condition monitoring under noisy real-world environments and domain shift conditions. Strong background noise remains a major challenge, as it degrades the reliability of detection systems. To address this issue, this paper proposes a noiseaware spectral subtraction method using dual-channel recordings, followed by a Conformer-based encoder for robust feature learning. Experimental results on the task dataset demonstrate the effectiveness of the proposed approach.

System characteristics
Classifier Conformer, kNN
System complexity 1314312
Acoustic features log-mel energies
Data augmentation mixup, SpecAugment
PDF

EAT-LORA AND NEAR/FAR STATISTICAL FEATURES FOR DCASE 2026 TASK 2 ANOMALOUS SOUND DETECTION

Baohe Zhu, LiLi Zhang, Ding Li
Research and Development Group, Shanghai Funda Acoustics Engineering Co., Ltd., Shanghai, China and Shanghai Funda Acoustics Engineering Co., Ltd., Shanghai, China

Abstract

This report describes the FDA submission to DCASE 2026 Task 2. The submitted systems are fixed -artifact, normal -only anomalous sound detectors designed for machine condition monitoring with two-channel recordings. Each system combines a handcrafted statistical branch using near/far acoustic information with a n Efficient Audio Transformer (EAT) embedding branch adapted by parameterefficient LoRA training. For each branch, train -normal examples are used to fit robust normalization, PCA, Mahalanobis, and k -nearestneighbor scoring artifacts. A domain -aware alpha adjustment, optional score clipping, and equal -weight branch fusion produce the final anomaly score. The binary decision threshold is the 0.99 quantile of train-normal fused scores and is not used for ranking metrics. On the development set, the best subm itted configuration obtains 62.83 percent Omega.

System characteristics
Classifier EAT LoRA, Mahalanobis distance, kNN, score fusion
System complexity 90318853, 90466309
Acoustic features log-mel energies and statistical audio features
Data augmentation frequency masking, gain scaling, time masking, time shifting
Decision making fixed score fusion and train-normal quantile threshold
System embeddings EAT
Subsystem count 2
External data usage pre-trained EAT model
Front end system two-channel near/far feature extraction
PDF