Faithfulness detection on In-domain Step-level Benchmark Knowledge
83.4FF1GeoFaith
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GeoFaith2026.05 | 83.4 | 70.3 | |
| GPT-o12026.05 | 81.2 | 72.3 | |
| DeepSeek-V32026.05 | 79.3 | 62.5 | |
| Qwen2.5-32B-Instruct2026.05 | 77 | 63.7 | |
| GPT-4o2026.05 | 76.1 | 59.8 | |
| Llama-3.1-70B-Instruct2026.05 | 75.8 | 52.2 | |
| o3-mini2026.05 | 74.8 | 57.6 | |
| LogicReward2026.05 | 73.5 | 53.2 | |
| FaithLens2026.05 | 69.7 | 42.3 | |
| HHEM2.12026.05 | 54.2 | 40.1 |