Event-level Video Understanding on MVBench
58.3Average AccuracyInternVL3.5-8B
Evaluation Results
| Method | Links | |
|---|---|---|
| InternVL3.5-8BBackbone=InternVL3.5-8B, Evaluation Protocol=APT-Tune2026.06 | 58.3 | |
| Qwen3-VL-8BBackbone=Qwen3-VL-8B, Evaluation Protocol=APT-Tune2026.06 | 57.6 | |
| InternVL3.5-8BBackbone=InternVL3.5-8B, Evaluation Protocol=Zero-shot2026.06 | 57.1 | |
| Qwen3-VL-8BBackbone=Qwen3-VL-8B, Evaluation Protocol=Zero-shot2026.06 | 56.8 | |
| InternVL3.5-4BBackbone=InternVL3.5-4B, Evaluation Protocol=APT-Tune2026.06 | 56.7 | |
| InternVL3.5-4BBackbone=InternVL3.5-4B, Evaluation Protocol=Zero-shot2026.06 | 55.8 | |
| InternVL3.5-2BBackbone=InternVL3.5-2B, Evaluation Protocol=APT-Tune2026.06 | 55.7 | |
| Qwen3-VL-4BBackbone=Qwen3-VL-4B, Evaluation Protocol=APT-Tune2026.06 | 55.2 | |
| Qwen3-VL-4BBackbone=Qwen3-VL-4B, Evaluation Protocol=Zero-shot2026.06 | 54.7 | |
| InternVL3.5-2BBackbone=InternVL3.5-2B, Evaluation Protocol=Zero-shot2026.06 | 54.5 | |
| Qwen3-VL-2BBackbone=Qwen3-VL-2B, Evaluation Protocol=APT-Tune2026.06 | 53.6 | |
| Qwen3-VL-2BBackbone=Qwen3-VL-2B, Evaluation Protocol=Zero-shot2026.06 | 51.4 |