Compositional Reasoning on GPQA
48.9AccuracyRCE
Evaluation Results
| Method | Links | |
|---|---|---|
| RCEModel=Qwen-14B2026.02 | 48.9 | |
| RCEModel=Llama-3-8B2026.02 | 43.1 | |
| RCEModel=Mistral-7B2026.02 | 41.4 | |
| BaseModel=Qwen-14B2026.02 | 36.7 | |
| DisCOModel=Mistral-7B2026.02 | 34.2 | |
| GRPOModel=Mistral-7B2026.02 | 32.4 | |
| ToTModel=Mistral-7B2026.02 | 31 | |
| SCModel=Mistral-7B, Samples (n)=162026.02 | 30.3 | |
| CoTModel=Mistral-7B2026.02 | 28.5 | |
| BaseModel=Llama-3-8B2026.02 | 27.3 | |
| BaseModel=Mistral-7B2026.02 | 24.1 |