List your company

Process Reward Modeling Evaluation on SCIPRM-Bench

48.58ChemBench Tool-Calling F1Sci-PRM
17.48425.55733.6341.703Jun 3, 2026

Evaluation results

MethodLinks
Sci-PRMModel Type=Open2026.0648.5853.6284.5756.1983.3376.9127.14
Gemini-3-FlashModel Type=Closed2026.0647.7159.7681.4455.8778.672.1122.73
GLM-4.6VModel Type=Closed2026.0647.1142.7559.7546.8985.473.0738.51
GPT-5-MiniModel Type=Closed2026.0646.349.8781.5653.3882.9475.9229.56
Qwen3-VL-8BModel Type=Open2026.0645.9547.9272.0949.5976.8568.1827.26
Doubao-Seed-1.6Model Type=Closed2026.0643.950.928051.2976.2168.5124.92
Qwen3-VL-32BModel Type=Open2026.0642.7343.4458.5945.6984.6975.3939
RandomModel Type=Baseline2026.0638.4533.4332.4235.2360.8353.2725.6
LLaMA-3.2-11B-VModel Type=Open2026.0618.6836.4446.2432.1357.5452.8125.41