Mathematical reasoning on MATH 500 (Pass@1, TC)
62.7Pass@1EAPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EAPOBackbone=Llama3.1-8B-Instruct2026.06 | 62.7 | 0.98 | |
| Reinforce++Backbone=Llama3.1-8B-Instruct2026.06 | 59.7 | 1.06 | |
| GRPOBackbone=Llama3.1-8B-Instruct2026.06 | 58.8 | 1.08 | |
| BaseBackbone=Llama3.1-8B-Instruct2026.06 | 47.2 | — | |
| TIRBackbone=Llama3.1-8B-Instruct2026.06 | 45.3 | 0.79 |