Reasoning on Reasoning Gym ARC-1D
30.67Success RateCBRL GRPO
Evaluation results
| Method | Links | |
|---|---|---|
| CBRL GRPOBackbone Model=Qwen2.5-3B-Instruct2026.03 | 30.67 | |
| GRPOBackbone Model=Qwen2.5-3B-Instruct2026.03 | 26 | |
| CBRL GRPOBackbone Model=Llama-3.2-3B-Instruct2026.03 | 25 | |
| GRPOBackbone Model=Llama-3.2-3B-Instruct2026.03 | 17 | |
| BaselineBackbone Model=Qwen2.5-3B-Instruct2026.03 | 4.67 | |
| Baseline w/ Few ShotsBackbone Model=Llama-3.2-3B-Instruct2026.03 | 3.67 | |
| BaselineBackbone Model=Llama-3.2-3B-Instruct2026.03 | 2.33 | |
| Baseline w/ Few ShotsBackbone Model=Qwen2.5-3B-Instruct2026.03 | 2 |