Geolocalization on MAPBench 1.0 (test-hard)
14.86Acc@500mThinking with Map + Reinforcement Learning + Parallelx4 & Verifier
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Thinking with Map + Reinforcement Learning + Parallelx4 & VerifierBackbone=Qwen3-VL-30B-A3B, Parallel Sampling=x4, Verifier=Yes2026.01 | 14.86 | 17.4 | 29.88 | 34.37 | 45.21 | 68.85 | |
| Thinking with Map + Reinforcement Learning + Parallelx2 & VerifierBackbone=Qwen3-VL-30B-A3B, Parallel Sampling=x2, Verifier=Yes2026.01 | 13.7 | 16.45 | 28.98 | 33.79 | 44.32 | 68.85 | |
| Thinking with Map + Reinforcement LearningBackbone=Qwen3-VL-30B-A3B, Training=RL2026.01 | 12.33 | 14.67 | 26.89 | 31.62 | 42.58 | 67.17 | |
| Thinking with MapBackbone=Qwen3-VL-30B-A3B2026.01 | 10.83 | 12.05 | 16.08 | 19.06 | 25.58 | 38.28 | |
| Gemini-3-ProModel Type=Closed Source, Tools=Google Search/Map2026.01 | 4.02 | 11.73 | 23.45 | 29.64 | 41.86 | 67.48 | |
| Qwen3-VL-235B-A22BModel Type=Open Source2026.01 | 0.63 | 3.42 | 13.41 | 19.31 | 32.88 | 57.18 | |
| Qwen3-VL-30B-A3BModel Type=Open Source2026.01 | 0.21 | 1.89 | 10.36 | 14.2 | 28.56 | 52.76 | |
| GPT-o3Model Type=Closed Source2026.01 | 0.05 | 0.74 | 4.53 | 9.1 | 20.73 | 44.13 | |
| GPT-5Model Type=Closed Source2026.01 | 0.05 | 0.79 | 4.1 | 8.94 | 22.3 | 47.45 | |
| GLOBE-7BModel Type=Open Source2026.01 | 0.05 | 0.85 | 6.34 | 11.35 | 27.68 | 52.29 | |
| GeoVista-7BModel Type=Open Source, Tools=Google Search2026.01 | 0 | 0.53 | 4.16 | 6.52 | 10.94 | 18.99 |