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

Multi-modal Task Evaluation on MM-MergeBench Unseen Tasks

79.3AVQA Score

RobustMerge

50.512857.986465.4672.9336Mar 6, 2026
Updated 4mo ago

Evaluation Results

MethodLinks
2026.03
79.345.799.2317.6237.99
2026.03
79.350.2410.8818.9439.84
2026.03
78.4844.8111.2113.9937.12
2026.03
77.2942.7211.1213.836.23
2026.03
76.3341.398.3418.236.06
2026.03
74.7837.377.5213.5733.31
2026.03
74.5736.287.8415.4433.53
2026.03
73.7537.677.5613.6233.15
2026.03
73.4338.447.4713.2333.14
2026.03
68.5641.5711.1611.833.27
2026.03
51.6228.275.9815.0125.22