Reason-intensive Grounding on Lisa Grounding
66.7P@0.5LaViDa-R1
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LaViDa-R1Model Category=Reinforcement Learning2026.02 | 66.7 | 60 | |
| VLM-R1Model Category=Reinforcement Learning2026.02 | 63.1 | — | |
| Qwen3-VL-8BModel Category=General-purpose VLMs2026.02 | 62.4 | 56.6 | |
| SegLLMModel Category=Specialist Models2026.02 | 61.3 | 55.2 | |
| LISA-7BModel Category=Specialist Models2026.02 | 49.4 | 50.6 | |
| LaViDa-O + SFTModel Category=Reinforcement Learning, Training Stage=Supervised Fine-Tuning2026.02 | 40.3 | 36.9 | |
| Qwen2.5-VL-7BModel Category=General-purpose VLMs2026.02 | 32 | 28.7 | |
| LaViDa-OModel Category=Reinforcement Learning2026.02 | 29.2 | 26.1 |