Captioning on Private dataset video + IMU
0.44ROUGE-LDoRA-adapted QwenVL-2.5
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DoRA-adapted QwenVL-2.5Size=7B2026.05 | 0.44 | 0.51 | |
| Molmo 2Size=8B2026.05 | 0.19 | 0.19 | |
| Sonnet 4.5Size=-2026.05 | 0.18 | 0.17 | |
| MiniCPM-V 4.5Size=8B2026.05 | 0.18 | 0.22 | |
| QwenVL 3.5Size=8B2026.05 | 0.18 | 0.18 | |
| QwenVL-2.5Size=7B2026.05 | 0.18 | 0.16 | |
| Nova ProSize=-2026.05 | 0.17 | 0.19 | |
| InternVL 3.5Size=8B2026.05 | 0.15 | 0.19 |