Multimodal Reasoning on GeoQA
49.2Mean@1Socratic-Solver-Geo (Stage3)
Evaluation Results
| Method | Links | |
|---|---|---|
| Socratic-Solver-Geo (Stage3)Data Scale=2.5k, Curriculum Stage=32026.02 | 49.2 | |
| Geo170k (G-LLaVA)Data Scale=10k2026.02 | 47.16 | |
| GeoReasoningData Scale=10k2026.02 | 46.76 | |
| KD (Our Synthesis)Data Scale=2.5k2026.02 | 46.58 | |
| R-CoTData Scale=7.2k2026.02 | 46.49 | |
| TrustGeoGenData Scale=10k2026.02 | 46.35 | |
| KD (Geo3K)Data Scale=3k2026.02 | 46.21 | |
| PGPS9KData Scale=10k2026.02 | 46.08 | |
| Socratic-Solver-Geo (Stage1)Data Scale=0.4k, Curriculum Stage=12026.02 | 45.14 | |
| Socratic-Solver-Geo (Stage2)Data Scale=1k, Curriculum Stage=22026.02 | 44.86 | |
| Qwen2.5-VL-7B-InstructMode=Zero-shot2026.02 | 43.92 |