Video Scene Description on WalkVLM (hard samples)
21.4ROUGE-LWalkVLM-LR
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| WalkVLM-LR2025.08 | 21.4 | 31.2 | 0.568 | |
| InternVL32025.08 | 17.1 | 25.9 | 0.39 | |
| DeepSeek-VL22025.08 | 16.2 | 26.8 | 0.443 | |
| Qwen2-VL2025.08 | 15.4 | 25.3 | 0.469 | |
| GPT-4o2025.08 | 6.3 | 22.6 | — |