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

Multimodal Long-Context Understanding on MileBench

51.5Object Existence

Full Attention

47.3448.4249.550.58Apr 18, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
51.52443.54215.814.740.533.14-
2026.04
51.524444215.815.93732.89-0.25
2026.04
5126.54341.516.215.341.533.570.43
2026.04
50.5244138.517.111.99.527.5-5.64
2026.04
48.532.551.552.515.214.84537.140.5
2026.04
47.531.55151.514.815.24536.64-
2026.04
47.53050.551.513.613.59.530.87-5.77
2026.04
47.531.55151.515.416.64436.790.15