Geometry Problem Solving on PyGeoX (Hard)
51AccuracyProprietary-C
Evaluation Results
| Method | Links | |
|---|---|---|
| Proprietary-CInference=reasoning, Evaluation Protocol=zero-shot2026.06 | 51 | |
| Qwen3-8B-RLInference=RL-trained, Evaluation Protocol=zero-shot2026.06 | 41 | |
| Proprietary-BInference=reasoning, Evaluation Protocol=zero-shot2026.06 | 38 | |
| Proprietary-AInference=default, Evaluation Protocol=zero-shot2026.06 | 37 | |
| DeepSeek-V3.2Inference=default, Evaluation Protocol=zero-shot2026.06 | 25 |