Scientific Reasoning on OlympiadBench Physics
87.3AccuracyHEART
Evaluation Results
| Method | Links | |
|---|---|---|
| HEARTModel=Deepseek-Reasoner2025.09 | 87.3 | |
| HEARTModel=Claude 4 Sonnet2025.09 | 86.67 | |
| HEARTModel=Gemini 2.5 Pro2025.09 | 84.44 | |
| HEARTModel=Gemini 2.5 Flash2025.09 | 84.13 | |
| HEARTModel=GPT-5 nano2025.09 | 83.81 | |
| VanillaModel=Claude 4 Sonnet2025.09 | 67.94 | |
| VanillaModel=Gemini 2.5 Flash2025.09 | 65.82 | |
| VanillaModel=GPT-5 nano2025.09 | 62.86 | |
| VanillaModel=Deepseek-Reasoner2025.09 | 60.63 | |
| VanillaModel=Gemini 2.5 Pro2025.09 | 60.11 | |
| HEARTModel=Gemma3 12b Instruct2025.09 | 53.02 | |
| HEARTModel=Gemma3 4b Instruct2025.09 | 25.19 | |
| VanillaModel=Gemma3 12b Instruct2025.09 | 22.96 | |
| VanillaModel=Gemma3 4b Instruct2025.09 | 8.78 |