Faithfulness detection on Step-level Benchmark In-domain Math
84.2FF1GeoFaith
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GeoFaith2026.05 | 84.2 | 73.1 | |
| GPT-o12026.05 | 83.5 | 72.8 | |
| GPT-4o2026.05 | 80.1 | 67.6 | |
| DeepSeek-V32026.05 | 79.8 | 69.1 | |
| Qwen2.5-32B-Instruct2026.05 | 78 | 62.5 | |
| o3-mini2026.05 | 76.7 | 61.4 | |
| Llama-3.1-70B-Instruct2026.05 | 73.5 | 52.8 | |
| LogicReward2026.05 | 72.2 | 59.8 | |
| FaithLens2026.05 | 61.2 | 49.7 | |
| HHEM2.12026.05 | 53.7 | 43.2 |