qiz1907 / LongLive-RAG

What it does

• First of its kind. Among open-ended AR long video generation methods, the first to formulate self-generated latent history as content-addressable retrieval memory.
• Plug-and-play. Works across Causal-Forcing, Self-Forcing, and LongLive with the base generator frozen.
• Searchable history. Retrieves the most relevant past latents as extra context for each new block.
• Window Temporal Delta Loss. Makes embeddings capture meaningful temporal change, not redundant local similarity.
• Consistent wins. Best average VBench-Long rank across lengths and backbones.