Geospatial region reasoning on EarthReason (test)
71.29Accuracy@0.5RemoteZero
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RemoteZerotraining_setup=Self-Evolution2026.05 | 71.29 | 61.7 | |
| RemoteReasonersupervision=supervised2026.05 | 68.11 | 69.29 | |
| RemoteReasoner2026.04 | 66.51 | 68.11 | |
| RemoteZerotraining_setup=External Teacher2026.05 | 65.05 | 57.95 | |
| RemoteAgent2026.04 | 57.81 | 54.22 | |
| Qwen2.5-VL-7B2026.05 | 45.82 | 41.8 | |
| Qwen2.5-VL-7B2026.04 | 41.21 | 45.82 | |
| VLM-R12026.05 | 33.31 | 29.44 | |
| DeepSeek-VL22026.05 | 12.67 | 18.62 | |
| DeepSeek-VL2-tiny2026.04 | 12.08 | 12.67 | |
| GeoChat2026.04 | 10.1 | 8.89 | |
| GeoChat2026.05 | 8.89 | 11.44 | |
| InternVL3.52026.05 | 5.26 | 6.52 |