List your company

Video Generation on VBench aesthetic and imaging quality dimensions

0.6499Aesthetic QualityLocalDPO
0.5931160.6078580.62260.637342Jan 7, 2026

Evaluation results

MethodLinks
LocalDPOBackbone=CogvideoX-2B, Fine-tuning strategy=Ours2026.010.64990.7080.273821.460.64922.23630.71734.90317.8568
LocalDPOBackbone=Wan 2.1-1.3B, Fine-tuning strategy=Ours2026.010.64160.64120.275421.420.72972.06520.54655.34717.9588
DenseDPOBackbone=Wan 2.1-1.3B, Fine-tuning strategy=DenseDPO2026.010.63750.63560.272821.370.68761.95190.54225.34317.8373
SFTBackbone=Wan 2.1-1.3B, Fine-tuning strategy=SFT2026.010.63730.63420.27321.380.7221.87790.51495.33557.7283
BaselineBackbone=Wan 2.1-1.3B, Fine-tuning strategy=Baseline2026.010.63630.62960.272721.370.68741.93870.54685.34447.83
Vanilla DPOBackbone=Wan 2.1-1.3B, Fine-tuning strategy=DPO2026.010.63530.63080.265421.370.59721.94370.52595.33837.8079
DenseDPOBackbone=CogvideoX-2B, Fine-tuning strategy=DenseDPO2026.010.63250.66060.265221.430.58842.16690.76754.88137.8157
Vanilla DPOBackbone=CogvideoX-2B, Fine-tuning strategy=DPO2026.010.63040.65980.265421.410.59722.18230.80674.79727.7862
SFTBackbone=CogvideoX-2B, Fine-tuning strategy=SFT2026.010.62930.65980.265921.470.55192.20030.74964.68197.6318
BaselineBackbone=CogvideoX-2B, Fine-tuning strategy=Baseline2026.010.62790.65890.265521.50.60792.1430.77414.87017.7871
LocalDPOBackbone=CogvideoX-5B, Fine-tuning strategy=Ours2026.010.62740.71070.278221.70.62974.51291.66824.111810.293
DenseDPOBackbone=CogvideoX-5B, Fine-tuning strategy=DenseDPO2026.010.62330.69620.267421.670.59593.32511.26714.98049.5726
SFTBackbone=CogvideoX-5B, Fine-tuning strategy=SFT2026.010.61320.6860.272821.580.57263.98691.41363.96199.3624
BaselineBackbone=CogvideoX-5B, Fine-tuning strategy=Baseline2026.010.6110.66310.269221.720.59574.16961.60053.9499.7191
Vanilla DPOBackbone=CogvideoX-5B, Fine-tuning strategy=DPO2026.010.59530.65340.265821.560.60124.08081.54983.96029.591