3-coloring on 3-coloring
98.56Forward Checking Success RateLlama-8b SFT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Llama-8b SFTSampling Strategy=BestOf82026.04 | 98.56 | 6.84 | 12.3 | 0.96 | |
| GPT5.2-Reasoning2026.04 | 95.25 | 24.33 | 12.38 | 2.08 | |
| Random2026.04 | 94.98 | — | — | 2.12 | |
| Llama-8b SFTSampling Strategy=BestOf12026.04 | 94.8 | 2.28 | 15.51 | 2.13 | |
| GPT5.22026.04 | 94.71 | 4.37 | 30.14 | 2.21 |