List your company

Reasoning Step Judgment on VisualProcessBench

74.1MMMU ScoreAthena-PRM
43.73251.61659.567.384Jun 11, 2025

Evaluation results

MethodLinks
Athena-PRMEvaluation Protocol=Multimodal PRMs, Model Size (Parameters)=7B2025.0674.161.260.772.773.865.9
Qwen2.5-VL-72BEvaluation Protocol=Open-source Models as Judge, Model Size (Parameters)=72B2025.0659.25959.762.962.360.5
Gemini-2.0-FlashEvaluation Protocol=Proprietary Models as Judge2025.0658.560.162.866.758.762.3
VisualPRM-8BEvaluation Protocol=Multimodal PRMs, Model Size (Parameters)=8B2025.0658.562.16162.761.862
GPT-4oEvaluation Protocol=Proprietary Models as Judge2025.0656.360.259.75963.360.3
GPT-4o-miniEvaluation Protocol=Proprietary Models as Judge2025.0653.658.957.156.758.557.9
Qwen2.5-VL-7BEvaluation Protocol=Open-source Models as Judge, Model Size (Parameters)=7B2025.0653.151.847.851.354.251
InternVL2.5-78BEvaluation Protocol=Open-source Models as Judge, Model Size (Parameters)=78B2025.065251.753.750.852.552.6
InternVL2.5-38BEvaluation Protocol=Open-source Models as Judge, Model Size (Parameters)=38B2025.0651.548.450.951.852.550.8
Random GuessingModel Category=Baseline2025.06505050505050
InternVL2.5-26BEvaluation Protocol=Open-source Models as Judge, Model Size (Parameters)=26B2025.0648.847.449.250.451.449.2
InternVL2.5-8BEvaluation Protocol=Open-source Models as Judge, Model Size (Parameters)=8B2025.0647.145.547.850.350.848
LLaVA-OV-72BEvaluation Protocol=Open-source Models as Judge, Model Size (Parameters)=72B2025.0646.148.4535757.352.3
LLaVA-OV-7BEvaluation Protocol=Open-source Models as Judge, Model Size (Parameters)=7B2025.0645.74342.244.752.544.4
MiniCPM-V2.6-8BEvaluation Protocol=Open-source Models as Judge, Model Size (Parameters)=8B2025.0644.950.958.946.757.450.4