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

Multimodal Language Model Unlearning on MLLMU-Bench 1.0 (test)

38.37Cloze Accuracy

RMU

0.597210.403620.2130.0164May 26, 2026
Updated 1d ago

Evaluation Results

MethodLinks
2026.05
38.379.590.267
2026.05
18.4934.590.186
2026.05
16.4438.110.226
2026.05
15.7538.380.236
2026.05
15.7537.570.175
2026.05
15.7534.860.237
2026.05
15.7537.290.224
2026.05
15.0738.920.22
2026.05
14.3835.680.203
2026.05
12.8452.970.297
2026.05
12.33400.277
2026.05
12.3332.970.228
2026.05
12.1653.780.321
2026.05
11.6435.140.229
2026.05
11.4954.60.301
2026.05
11.4954.320.304
2026.05
11.4954.050.296
2026.05
11.4540.410.244
2026.05
10.9643.510.242
2026.05
10.27400.241
2026.05
10.2739.190.264
2026.05
10.1454.60.316
2026.05
10.1453.780.321
2026.05
9.5928.380.213
2026.05
9.5838.110.269
2026.05
9.4651.620.24
2026.05
9.4652.70.237
2026.05
9.4651.080.225
2026.05
8.2240.270.271
2026.05
2.0536.470.199