Cross-view Correspondence on uCO3D ReasonMatch-Bench high-divergence samples
72.2PrecisionHuman
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| HumanAnnotator=Human2026.06 | 72.2 | 56.6 | 62.1 | |
| DCRLModel Category=Open-Source Models2026.06 | 25.9 | 30.4 | 27.8 | |
| GPT-5-miniModel Category=Closed-Source Models2026.06 | 23.6 | 29.1 | 25.8 | |
| Claude-4.5-SonnetModel Category=Closed-Source Models2026.06 | 17.4 | 21.6 | 19.2 | |
| Gemini-2.5-ProModel Category=Closed-Source Models2026.06 | 17.1 | 19.3 | 18 | |
| Qwen3-VL-235BModel Category=Open-Source Models2026.06 | 17.1 | 20.9 | 18.7 |