Video Generation on GenAI-Bench
82.5AccuracyMSRL + voting@16
Evaluation Results
| Method | Links | |
|---|---|---|
| MSRL + voting@16Params.=8B, Training Preference Data=InternVL3.5-8B, Evaluation Strategy=voting@162026.03 | 82.5 | |
| UnifiedReward-think (w/ CoT)Params.=7B, Evaluation Strategy=w/ CoT2026.03 | 82.3 | |
| MSRL + voting@16Params.=4B, Training Preference Data=InternVL3.5-4B, Evaluation Strategy=voting@162026.03 | 81.8 | |
| MSRLParams.=8B, Training Preference Data=InternVL3.5-8B2026.03 | 81.4 | |
| MSRLParams.=4B, Training Preference Data=InternVL3.5-4B2026.03 | 80.6 | |
| UnifiedRewardParams.=7B2026.03 | 77.2 | |
| Discriminative MRMParams.=4B, Training Preference Data=InternVL3.5-4B2026.03 | 76.4 | |
| Generative MRMParams.=4B, Training Preference Data=InternVL3.5-4B2026.03 | 75.9 | |
| VisionRewardParams.=19B2026.03 | 73.1 | |
| Discriminative MRMParams.=8B, Training Preference Data=InternVL3.5-8B2026.03 | 70.3 | |
| Generative MRMParams.=8B, Training Preference Data=InternVL3.5-8B2026.03 | 68.3 | |
| InternVL3.5-14BParams.=14B2026.03 | 66.5 | |
| GPT-4oParams.=-2026.03 | 63.5 | |
| InternVL3.5-8BParams.=8B2026.03 | 60.1 |