HACKOBAR_item
[GH]score: 0.29

A²RD: Agentic Autoregressive Diffusion for Long Video Generation

May 12, 2026
A²RD proposes an agentic autoregressive diffusion framework for long video generation using iterative retrieval, synthesis, refinement, and memory updates. Targeting coherence over extended temporal sequences, this addresses a key limitation in current video diffusion models. Relevant for researchers in video generation and agentic AI pipelines.