Scene-Graph Grounding and Editing on Caption-Free Scene-Graph Grounding and Editing
78.7IoUQwen2.5-VL-7B (Ours)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen2.5-VL-7B (Ours)variant=Ours2026.03 | 78.7 | 97.9 | |
| Qwen2.5-VL-7B (CoT-SFT)variant=CoT-SFT2026.03 | 64.1 | 82.9 | |
| Qwen2.5-VL-7B (Ans-SFT)variant=Ans-SFT2026.03 | 52.1 | 70 | |
| Qwen3-VL-235B-Thinkingvariant=Thinking2026.03 | 47.2 | 74.4 | |
| Qwen3-VL-235B-Instructvariant=Instruct2026.03 | 40.8 | 63.1 | |
| Qwen3-8B (Ours, for reference)variant=Ours, for reference2026.03 | 20.9 | 98.4 |