ResearchBenchmarksVideo Temporal Consistency on VBench Cross-ChunkFollow99.03DINO SimilaritySTAS98.010898.275498.5498.8046Mar 18, 2026Evaluation ResultsMethodMethodLinksDINO SimilarityCLIP SimilaritySTASBackbone Model=Wan2.1-...Backbone Model=Wan2.1-1.3B2026.0399.0398.66STASBackbone Model=Wan2.2-5BBackbone Model=Wan2.2-5B2026.0399.0398.95VanillaBackbone Model=Wan2.2-5BBackbone Model=Wan2.2-5B2026.039998.94VanillaBackbone Model=Wan2.1-...Backbone Model=Wan2.1-1.3B2026.0398.9498.58STASBackbone Model=CogVide...Backbone Model=CogVideoX-5B2026.0398.1698.15VanillaBackbone Model=CogVide...Backbone Model=CogVideoX-5B2026.0398.0598.05