Motion Understanding on Robotics
38.2AccuracyNVILA-Video-8B
Evaluation Results
| Method | Links | |
|---|---|---|
| NVILA-Video-8BMode=Fine-tuned, Data=FoundationMotion2025.12 | 38.2 | |
| Qwen-2.5-VL-72BMode=Zero-shot2025.12 | 36.7 | |
| NVILA-Video-15BMode=Fine-tuned, Data=FoundationMotion2025.12 | 36.3 | |
| Gemini-2.5-FlashMode=Zero-shot2025.12 | 36.1 | |
| Qwen-2.5-VL-7BMode=Fine-tuned, Data=FoundationMotion2025.12 | 32.5 | |
| Qwen-2.5-VL-7BMode=Zero-shot2025.12 | 28.3 | |
| NVILA-Video-15BMode=Fine-tuned, Data=PLM 467k2025.12 | 27.4 | |
| NVILA-Video-8BMode=Fine-tuned, Data=PLM 467k2025.12 | 26.5 | |
| NVILA-Video-15BMode=Zero-shot2025.12 | 21.4 | |
| NVILA-Video-8BMode=Zero-shot2025.12 | 20.4 |