Mathematical Reasoning on OlympiadBench (Pass@1, Avg. Pass@1)
45.2Pass@1LILO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LILOModel=Qwen2.5-Math-7B, Avg. Rollouts=4.0x2026.02 | 45.2 | 40.1 | |
| PACED-RLModel=Qwen2.5-Math-7B, Avg. Rollouts=1.0x2026.02 | 44.6 | 40.1 | |
| DSModel=Qwen2.5-Math-7B, Avg. Rollouts=2.5x2026.02 | 44 | 39.4 | |
| MoPPSModel=Qwen2.5-Math-7B, Avg. Rollouts=1.0x2026.02 | 42.2 | 38.6 | |
| GRPOModel=Qwen2.5-Math-7B, Avg. Rollouts=1.0x2026.02 | 39.7 | 37.2 | |
| FlowRLModel=Qwen2.5-Math-7B, Avg. Rollouts=1.0x2026.02 | 39.6 | 36.5 | |
| PACED-RLModel=Qwen2.5-Math-1.5B, Avg. Rollouts=1.0x2026.02 | 37.9 | 32.1 | |
| DSModel=Qwen2.5-Math-1.5B, Avg. Rollouts=2.0x2026.02 | 36.8 | 31.2 | |
| MoPPSModel=Qwen2.5-Math-1.5B, Avg. Rollouts=1.0x2026.02 | 36.6 | 30.4 | |
| LILOModel=Qwen2.5-Math-1.5B, Avg. Rollouts=4.0x2026.02 | 36.2 | 31.2 | |
| GRPOModel=Qwen2.5-Math-1.5B, Avg. Rollouts=1.0x2026.02 | 34.2 | 29.1 | |
| FlowRLModel=Qwen2.5-Math-1.5B, Avg. Rollouts=1.0x2026.02 | 32.4 | 27.1 |