List your company

Multimodal Evaluation on MM (test)

51.6Rubric RewardPOW3R dynamic
33.19237.97142.7547.529May 19, 2026

Evaluation results

MethodLinks
POW3R dynamicBase policy=Qwen3-VL-8B2026.0551.622.6
Category-balancedBase policy=Qwen3-VL-8B2026.0550.421.6
Static scalarBase policy=Qwen3-VL-8B2026.0549.520.7
POW3R dynamicBase policy=Qwen3-VL-4B2026.0548.820.2
Category-balancedBase policy=Qwen3-VL-4B2026.0547.918.7
Static scalarBase policy=Qwen3-VL-4B2026.0547.117.9
BinaryBase policy=Qwen3-VL-8B2026.054718.3
BaseBase policy=Qwen3-VL-8B2026.0546.818.4
BinaryBase policy=Qwen3-VL-4B2026.0542.814.6
BaseBase policy=Qwen3-VL-4B2026.0542.614.7
POW3R dynamicBase policy=Gemma3-4B2026.0537.810.3
Category-balancedBase policy=Gemma3-4B2026.0536.79.4
Static scalarBase policy=Gemma3-4B2026.0535.88.5
BaseBase policy=Gemma3-4B2026.0534.17.7
BinaryBase policy=Gemma3-4B2026.0533.97.6