AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Unlocking Any-Order Generation in Pretrained Autoregressive Image Models

Rishav Pramanik, Marco Pedersoli, Zhaozheng Yin

NeurIPS 2026 · Image generation › Autoregressive and latent generation · approach: Accelerating autoregressive image generation · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a method to unlock arbitrary-order (any-order) generation in already-pretrained raster-scan autoregressive image models, enabling inpainting, outpainting, and editing at inference time.

Problem

Raster-scan autoregressive image generators are fixed to a single factorization of the image distribution, so they cannot inpaint, outpaint, or edit at inference; existing any-order solutions (including discrete diffusion) train from scratch on mixed token orders at roughly 3x compute.

Approach

Adapts existing pretrained raster-scan AR image models to any-order generation rather than training from scratch on mixed orderings, avoiding the ~3x training compute cost of prior any-order approaches. Details are not available in the provided abstract.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
adapt pretrained raster-scan checkpoint rather than train random init at matched total epochs (ablation)any-order recipe on LlamaGen-L · ImageNet class-conditionalFID: 5.60 -> 4.24Improved
adapt pretrained raster-scan checkpoint rather than train random init at matched total epochs (ablation)any-order recipe on LlamaGen-XL · ImageNet class-conditionalFID: 4.81 -> 3.91Improved
add ordering curriculum (linear alpha ramp) instead of alpha = 1 from step 0 (ablation)any-order adaptation of LlamaGen-L · ImageNet class-conditionalFID: 5.86 -> 4.24Improved
add future-aware positional embeddings (ablation)any-order adaptation · training losstraining loss convergenceImproved
train embedding table at 10x base learning rate (ablation)future-aware embedding table · any-order adaptationconvergence and any-order qualityImproved
add KL regularization toward frozen raster-scan teacher (ablation)any-order adaptation of LlamaGen-L · ImageNet class-conditionalFID: 4.24 -> 9.62Hurt
inject future-aware embedding at every layer (AdaLN-style) instead of input layer only (ablation)future-aware embedding · design choicequalityNo clear effect
adapt pretrained raster model to any-orderLlamaGen-XL · ImageNet 256 class-conditionalFID: 2.63 -> 3.91Hurt
adapt pretrained raster model to any-orderRAR-L · ImageNet 256 class-conditionalFID: 1.70 -> 3.63Hurt
use any-order adapted model instead of MaskGIT (cosine) for inpaintingImageNet inpainting · ImageNet inpaintingFID: 10.44 -> 6.21Improved
use Halton schedule for MaskGIT instead of cosineMaskGIT baseline · ImageNet inpaintingFID: 10.44 -> 7.82Improved
adapted model versus MAR-L on inpainting and outpaintingImageNet inpainting/outpainting · ImageNetFID: 6.63 -> 6.21 inpainting; 6.86 -> 5.84 outpaintingNo clear effect

Limitations

Experimental setup

Baselines: discrete diffusion models

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy