Reasoning on Reasoning Gym Spell Backward
97Success RateCBRL GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| CBRL GRPOBackbone Model=Llama-3.2-3B-Instruct2026.03 | 97 | |
| GRPOBackbone Model=Llama-3.2-3B-Instruct2026.03 | 95.67 | |
| CBRL GRPOBackbone Model=Qwen2.5-3B-Instruct2026.03 | 52.67 | |
| GRPOBackbone Model=Qwen2.5-3B-Instruct2026.03 | 51 | |
| Baseline w/ Few ShotsBackbone Model=Llama-3.2-3B-Instruct2026.03 | 23.67 | |
| BaselineBackbone Model=Llama-3.2-3B-Instruct2026.03 | 20 | |
| BaselineBackbone Model=Qwen2.5-3B-Instruct2026.03 | 4.33 | |
| Baseline w/ Few ShotsBackbone Model=Qwen2.5-3B-Instruct2026.03 | 4 |