Explanation Refinement on FOLIO
85.25Initial ScoreLLM-TP Tree
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LLM-TP TreeBackbone=Qwen3-max2026.01 | 85.25 | 95.08 | |
| LLM-TP TreeBackbone=Deepseek-V3.12026.01 | 83.61 | 92.62 | |
| LLM-TP TreeBackbone=GPT-4o2026.01 | 81.15 | 90.16 | |
| Faithful-RefinerBackbone=Deepseek-V3.12026.01 | 71.31 | 81.15 | |
| Faithful-RefinerBackbone=Qwen3-max2026.01 | 69.67 | 84.43 | |
| Faithful-RefinerBackbone=GPT-4o2026.01 | 68.03 | 78.69 | |
| LLM-TP TreeBackbone=GPT-5 nano2026.01 | 66.39 | 81.15 | |
| LLM-TP TreeBackbone=Grok-4 fast2026.01 | 62.3 | 79.51 | |
| Explanation-RefinerBackbone=Qwen3-max2026.01 | 55.74 | 68.85 | |
| Faithful-RefinerBackbone=Grok-4 fast2026.01 | 54.92 | 65.57 | |
| Explanation-RefinerBackbone=GPT-4o2026.01 | 53.28 | 63.93 | |
| Faithful-RefinerBackbone=GPT-5 nano2026.01 | 53.28 | 68.03 | |
| Explanation-RefinerBackbone=Deepseek-V3.12026.01 | 50.82 | 62.3 | |
| Explanation-RefinerBackbone=GPT-5 nano2026.01 | 46.72 | 59.84 | |
| Explanation-RefinerBackbone=Grok-4 fast2026.01 | 45.08 | 52.46 |