Geolocation on GeoBrowse Level 2 BrowseComp-style
26.3City AccuracyGATE
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GATEBackbone=Gemini-3-Pro2026.04 | 26.3 | 36.8 | 34.9 | 34.7 | |
| GPT-5.2Framework=Direct Inference2026.04 | 21.1 | 21.1 | 20.6 | 20.8 | |
| Gemini-3-ProFramework=Direct Inference2026.04 | 21.1 | 21.1 | 23.8 | 22.8 | |
| Claude-4.5-OpusFramework=Direct Inference2026.04 | 21.1 | 26.3 | 27 | 25.7 | |
| GATEBackbone=GPT-52026.04 | 21.1 | 31.6 | 31.8 | 30.7 | |
| GATEBackbone=Claude-4.5-Opus2026.04 | 21.1 | 36.8 | 33.3 | 32.7 | |
| GPT-5Framework=Direct Inference2026.04 | 15.8 | 10.8 | 20.6 | 17.8 | |
| OmniSearchFramework=Open Source Agents2026.04 | 15.8 | 21.1 | 19 | 18.8 | |
| WebWatcherFramework=Open Source Agents2026.04 | 15.8 | 26.3 | 20.6 | 20.8 | |
| GATEBackbone=GPT-4o2026.04 | 15.8 | 21.1 | 22.2 | 21.8 | |
| GATEBackbone=GPT-4.12026.04 | 15.8 | 21.1 | 23.8 | 22.8 | |
| GATEBackbone=Gemini-2.5-Pro2026.04 | 15.8 | 36.8 | 30.2 | 29.7 | |
| Gemini-2.5-ProFramework=Direct Inference2026.04 | 10.8 | 10.8 | 17.5 | 14.8 | |
| GPT-4oFramework=Direct Inference2026.04 | 10.5 | 15.8 | 11.1 | 11.9 | |
| PyvisionFramework=Open Source Agents2026.04 | 10.5 | 15.8 | 14.3 | 13.9 | |
| GPT-4.1Framework=Direct Inference2026.04 | 5.3 | 5.3 | 15.9 | 11.9 | |
| Gemini-2.5-flashFramework=Direct Inference2026.04 | 5.3 | 10.8 | 9.9 | 12.9 | |
| Llama-3.2-90BFramework=Direct Inference2026.04 | 5.3 | 5.3 | 9.5 | 7.9 | |
| GATEBackbone=Llama-3.2-90B2026.04 | 5.3 | 10.5 | 12.7 | 11.9 | |
| GATEBackbone=Qwen-3-VL-32B2026.04 | 5.3 | 5.3 | 9.5 | 8.9 | |
| Qwen-3-VL-32BFramework=Direct Inference2026.04 | 0 | 5.3 | 9.5 | 6.9 | |
| Qwen-2.5-VL-7BFramework=Direct Inference2026.04 | 0 | 0 | 3.2 | 1.9 | |
| Qwen-2.5-VL-32BFramework=Direct Inference2026.04 | 0 | 0 | 6.3 | 4 | |
| GATEBackbone=Qwen-2.5-VL-7B2026.04 | 0 | 5.3 | 6.4 | 4.9 |