List your company

Scientific Reasoning on OlympiadBench Physics

87.3AccuracyHEART
5.639226.839648.0469.2404Sep 26, 2025

Evaluation results

MethodLinks
HEARTModel=Deepseek-Reasoner2025.0987.3
HEARTModel=Claude 4 Sonnet2025.0986.67
HEARTModel=Gemini 2.5 Pro2025.0984.44
HEARTModel=Gemini 2.5 Flash2025.0984.13
HEARTModel=GPT-5 nano2025.0983.81
VanillaModel=Claude 4 Sonnet2025.0967.94
VanillaModel=Gemini 2.5 Flash2025.0965.82
VanillaModel=GPT-5 nano2025.0962.86
VanillaModel=Deepseek-Reasoner2025.0960.63
VanillaModel=Gemini 2.5 Pro2025.0960.11
HEARTModel=Gemma3 12b Instruct2025.0953.02
HEARTModel=Gemma3 4b Instruct2025.0925.19
VanillaModel=Gemma3 12b Instruct2025.0922.96
VanillaModel=Gemma3 4b Instruct2025.098.78