Compositional Reasoning on HLE
23.1AccuracyRCE
Evaluation Results
| Method | Links | |
|---|---|---|
| RCEModel=Qwen-14B2026.02 | 23.1 | |
| RCEModel=Llama-3-8B2026.02 | 20.2 | |
| RCEModel=Mistral-7B2026.02 | 18.7 | |
| BaseModel=Qwen-14B2026.02 | 14.3 | |
| DisCOModel=Mistral-7B2026.02 | 13.8 | |
| GRPOModel=Mistral-7B2026.02 | 12.6 | |
| ToTModel=Mistral-7B2026.02 | 11.9 | |
| SCModel=Mistral-7B, Samples (n)=162026.02 | 11.4 | |
| CoTModel=Mistral-7B2026.02 | 10.1 | |
| BaseModel=Llama-3-8B2026.02 | 9.6 | |
| BaseModel=Mistral-7B2026.02 | 8.2 |