Faithfulness detection on In-domain Step-level Benchmark Reasoning
84.5FF1GeoFaith
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GeoFaith2026.05 | 84.5 | 68.8 | |
| GPT-o12026.05 | 83.8 | 72.5 | |
| DeepSeek-V32026.05 | 82.6 | 67.2 | |
| GPT-4o2026.05 | 82.4 | 62.8 | |
| Qwen2.5-32B-Instruct2026.05 | 80.1 | 62.3 | |
| LogicReward2026.05 | 80 | 67.1 | |
| o3-mini2026.05 | 78.9 | 59.8 | |
| Llama-3.1-70B-Instruct2026.05 | 72.3 | 59.7 | |
| FaithLens2026.05 | 71.2 | 52.3 | |
| HHEM2.12026.05 | 59.2 | 39.7 |