Video Generation on VideoAlign
8.47VQ ScoreDMD2
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DMD22026.05 | 8.47 | 4.15 | 6.24 | 18.86 | |
| SGMD2026.05 | 8.19 | 4.99 | 6.19 | 19.36 | |
| AR-CoPO (on-policy)Base Model=Self-Forcing, Alignment Strategy=on-policy2026.03 | 4.15 | 2.06 | 2.3 | 8.51 | |
| AR-CoPO (merged)Base Model=Self-Forcing, Alignment Strategy=semi-on-policy and on-policy merged2026.03 | 4 | 1.86 | 2.36 | 8.22 | |
| Causal-Forcing2026.03 | 3.97 | 1.43 | 2.4 | 7.79 | |
| LongLive2026.03 | 3.87 | 1.76 | 2.43 | 8.06 | |
| Self-Forcing2026.03 | 3.8 | 1.68 | 2.28 | 7.76 | |
| AR-CoPO (semi)Base Model=Self-Forcing, Alignment Strategy=semi-on-policy2026.03 | 3.7 | 1.6 | 2.3 | 7.61 | |
| TQDModel=Wan 1.3B, Setting=Set-C2026.03 | 3.3743 | 2.22 | — | — | |
| TQDModel=Wan 1.3B, Setting=Set-A2026.03 | 3.345 | 2.2557 | — | — | |
| BaselineModel=Wan 1.3B, Setting=Set-C2026.03 | 3.3378 | 2.1917 | — | — | |
| BasemodelModel=Wan 1.3B, Setting=-2026.03 | 3.3236 | 2.1993 | — | — | |
| TQDModel=Wan 1.3B, Setting=Set-B2026.03 | 3.2679 | 2.1477 | — | — | |
| BaselineModel=CogVideoX 5B, Setting=Set-A2026.03 | 3.2593 | 2.1855 | — | — | |
| TQDModel=CogVideoX 5B, Setting=Set-A2026.03 | 3.257 | 2.2345 | — | — | |
| BaselineModel=Wan 1.3B, Setting=Set-A2026.03 | 3.2537 | 2.1388 | — | — | |
| TQDModel=CogVideoX 5B, Setting=Set-B2026.03 | 3.2499 | 2.201 | — | — | |
| BaselineModel=CogVideoX 5B, Setting=Set-B2026.03 | 3.2407 | 2.1821 | — | — | |
| TQDModel=CogVideoX 5B, Setting=Set-C2026.03 | 3.2392 | 2.2448 | — | — | |
| BasemodelModel=CogVideoX 5B, Setting=-2026.03 | 3.2379 | 2.2134 | — | — | |
| BaselineModel=CogVideoX 5B, Setting=Set-C2026.03 | 3.2373 | 2.2028 | — | — | |
| BaselineModel=Wan 1.3B, Setting=Set-B2026.03 | 3.2338 | 2.0905 | — | — | |
| TGO-LoRABackbone=Wan 1.3B, Reward Model=VideoReward2026.05 | -0.5631 | 0.0627 | -0.6753 | -1.1757 | |
| KTO-LoRABackbone=Wan 1.3B, Reward Model=VideoReward2026.05 | -0.5832 | -0.1025 | -0.6846 | -1.3703 | |
| SFT-LoRABackbone=Wan 1.3B, Reward Model=VideoReward2026.05 | -0.6054 | -0.4159 | -0.6705 | -1.6918 | |
| OriginalBackbone=Wan 1.3B, Reward Model=VideoReward2026.05 | -0.7963 | -0.4316 | -0.8639 | -2.0918 |