Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Multimodal Language Model Unlearning on MLLMU-Bench forget set 1.0

35.14Cloze Accuracy

RMU

3.57611.770519.96528.1595May 26, 2026
Updated 1d ago

Evaluation Results

MethodLinks
2026.05
35.1413.010.472
2026.05
28.76400.245
2026.05
27.3947.840.337
2026.05
25.3450.540.486
2026.05
23.9749.190.517
2026.05
18.4950.810.472
2026.05
18.4945.680.284
2026.05
17.8152.160.506
2026.05
14.5841.630.53
2026.05
13.6939.650.465
2026.05
13.6940.540.491
2026.05
13.5152.430.504
2026.05
13.5152.430.506
2026.05
13.5151.080.472
2026.05
13.5153.240.491
2026.05
13.5151.890.461
2026.05
13.5153.240.499
2026.05
12.8448.920.39
2026.05
12.8450.540.359
2026.05
12.8448.920.351
2026.05
12.3332.420.473
2026.05
12.3335.140.429
2026.05
12.3332.160.485
2026.05
11.4956.760.461
2026.05
10.9533.510.343
2026.05
10.2848.920.428
2026.05
10.2747.830.241
2026.05
10.2729.460.494
2026.05
6.1736.760.374
2026.05
4.7915.680.251