Cross-view Correspondence on DL3DV ReasonMatch-Bench high-divergence samples
94.3PrecisionHuman
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| HumanAnnotator=Human2026.06 | 94.3 | 93.3 | 93.5 | |
| DCRLModel Category=Open-Source Models2026.06 | 53.3 | 63.2 | 57.7 | |
| GPT-5-miniModel Category=Closed-Source Models2026.06 | 32.5 | 40.4 | 35.9 | |
| Gemini-2.5-ProModel Category=Closed-Source Models2026.06 | 25.1 | 28.3 | 26.5 | |
| Qwen3-VL-235BModel Category=Open-Source Models2026.06 | 23.2 | 28.1 | 25.3 | |
| Claude-4.5-SonnetModel Category=Closed-Source Models2026.06 | 20.6 | 24.8 | 22.2 |