Wide-Baseline Correspondence Reasoning on ReasonMatch-Bench DL3DV high-divergence
93.5F1 ScoreHuman
Evaluation Results
| Method | Links | |
|---|---|---|
| Human2026.06 | 93.5 | |
| DCRL2026.06 | 57.7 | |
| GPT-5-mini2026.06 | 35.9 | |
| Gemini-2.5-Pro2026.06 | 26.5 | |
| Qwen3-VL-235B2026.06 | 25.3 | |
| Claude-4.5-Sonnet2026.06 | 22.2 |