Infeasibility Detection on 3-coloring problem (test)
89.2Infeasibility Detection AccuracyLlama3.1-8b SFT BestOf8
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Llama3.1-8b SFT BestOf8SFT=true, BestOf=82026.04 | 89.2 | 95.82 | 81.86 | 16.55 | |
| Llama3.1-8b SFT BestOf1SFT=true, BestOf=12026.04 | 83.1 | 83.84 | 82.28 | 11.22 | |
| GPT5.2-ReasoningReasoning=true2026.04 | 62.3 | 28.33 | 100 | 42.98 | |
| Grok-4.1-Fast (Reasoning)Reasoning=true2026.04 | 60 | 23.5 | 99.2 | 111.42 | |
| GPT5.22026.04 | 50.4 | 5.89 | 99.79 | 7.92 | |
| Deepseek-V3.22026.04 | 48.1 | 26.9 | 71.7 | 11.75 | |
| Llama3.3-70B2026.04 | 47.7 | 1.3 | 99.2 | 1.08 |