Drift-AR enables single-step (1-NFE) high-fidelity image generation by reinterpreting AR prediction entropy as a physical drifting field.
March 31, 2026
Original Paper
Drift-AR: Single-Step Visual Autoregressive Generation via Anti-Symmetric Drifting
arXiv · 2603.28049
The Takeaway
It resolves the speed bottleneck of AR-Diffusion hybrids by using entropy to govern a continuous drift toward the data manifold. This allows for real-time visual autoregressive generation without the need for iterative denoising or multi-step sampling.
From the abstract
Autoregressive (AR)-Diffusion hybrid paradigms combine AR's structured semantic modeling with diffusion's high-fidelity synthesis, yet suffer from a dual speed bottleneck: the sequential AR stage and the iterative multi-step denoising of the diffusion vision decode stage. Existing methods address each in isolation without a unified principle design. We observe that the per-position \emph{prediction entropy} of continuous-space AR models naturally encodes spatially varying generation uncertainty,