A$^2$RD: Agentic Autoregressive Diffusion for Long Video Consistency

TL;DR

A$^2$RD is a novel agentic autoregressive diffusion framework that enhances long video synthesis by self-improvement and consistency mechanisms, outperforming existing methods on new benchmarks.

Contribution

The paper introduces A$^2$RD, a new architecture for long video synthesis that decouples creativity from consistency, with a self-improving cycle and a challenging benchmark LVBench-C.

Findings

A$^2$RD outperforms baselines by up to 30% in consistency.

Achieves 20% improvement in narrative coherence.

Human evaluations show better motion and transition smoothness.

Abstract

Synthesizing consistent and coherent long video remains a fundamental challenge. Existing methods suffer from semantic drift and narrative collapse over long horizons. We present A$^2$RD, an Agentic Auto-Regressive Diffusion architecture that decouples creative synthesis from consistency enforcement. A$^2$RD formulates long video synthesis as a closed-loop process that synthesizes and self-improves video segment-by-segment through a Retrieve--Synthesize--Refine--Update cycle. It comprises three core components: (i) Multimodal Video Memory that tracks video progression across modalities; (ii) Adaptive Segment Generation that switches among generation modes for natural progression and visual consistency; and (iii) Hierarchical Test-Time Self-Improvement that self-improves each segment at frame and video levels to prevent error propagation. We further introduce LVBench-C, a challenging benchmark with non-linear entity and environment transitions to stress-test long-horizon consistency. Across public and LVBench-C benchmarks spanning one- to ten-minute videos, A$^2$RD outperforms state-of-the-art baselines by up to 30% in consistency and 20% in narrative coherence. Human evaluations corroborate these gains while also highlighting notable improvements in motion and transition smoothness.