REVIEW 11 cited by
Beyond Next-Token: Next-X Prediction for Autoregressive Visual Generation
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
abstract
Autoregressive (AR) modeling, known for its next-token prediction paradigm, underpins state-of-the-art language and visual generative models. Traditionally, a ``token'' is treated as the smallest prediction unit, often a discrete symbol in language or a quantized patch in vision. However, the optimal token definition for 2D image structures remains an open question. Moreover, AR models suffer from exposure bias, where teacher forcing during training leads to error accumulation at inference. In this paper, we propose xAR, a generalized AR framework that extends the notion of a token to an entity X, which can represent an individual patch token, a cell (a $k\times k$ grouping of neighboring patches), a subsample (a non-local grouping of distant patches), a scale (coarse-to-fine resolution), or even a whole image. Additionally, we reformulate discrete token classification as continuous entity regression, leveraging flow-matching methods at each AR step. This approach conditions training on noisy entities instead of ground truth tokens, leading to Noisy Context Learning, which effectively alleviates exposure bias. As a result, xAR offers two key advantages: (1) it enables flexible prediction units that capture different contextual granularity and spatial structures, and (2) it mitigates exposure bias by avoiding reliance on teacher forcing. On ImageNet-256 generation benchmark, our base model, xAR-B (172M), outperforms DiT-XL/SiT-XL (675M) while achieving 20$\times$ faster inference. Meanwhile, xAR-H sets a new state-of-the-art with an FID of 1.24, running 2.2$\times$ faster than the previous best-performing model without relying on vision foundation modules (e.g., DINOv2) or advanced guidance interval sampling.
Forward citations
Cited by 11 Pith papers
-
D-AR: Diffusion via Autoregressive Models
D-AR recasts pixel-space diffusion as vanilla autoregressive next-token prediction using a diffusion-ordered discrete tokenizer, reaching 2.09 FID on ImageNet 256x256 with a 775M Llama backbone.
-
Revisiting Autoregressive Models for Generative Image Classification
Order-marginalized any-order AR models (RandAR) outperform diffusion generative classifiers on ImageNet and OOD sets and match strong SSL models at far lower cost.
-
MIMFlow: Integrating Masked Image Modeling with Normalizing Flows for End-to-End Image Generation
MIMFlow uses a VAE on masked images to feed semantic latents to a normalizing flow while a decoder handles high-frequency details, reporting FID 2.50 and 71.3% linear probing on ImageNet 256x256 with 128 tokens.
-
SRC-Flow: Compact Semantic Representations Enable Normalizing Flows for Image Generation
SRC-Flow compresses RAE features via a Semantic Representation Compressor into a low-dimensional space, enabling normalizing flows to reach gFID 1.65 on ImageNet 256x256 and 2.07 on 512x512 while retaining exact likelihoods.
-
Progressive Checkerboards for Autoregressive Multiscale Image Generation
A balanced multiscale checkerboard sampling order for autoregressive image generation allows large scale-up factors without quality loss, because only the total number of serial steps matters.
-
MixFlow Training: Alleviating Exposure Bias with Slowed Interpolation Mixture
Training diffusion models on a mixture of higher-noise interpolations (MixFlow) improves generation FID across SiT, REPA, RAE and SD3.5, reaching ImageNet 256 gFID 1.43 after post-training.
-
UniCode$^2$: Cascaded Large-scale Codebooks for Unified Multimodal Understanding and Generation
UniCode² builds a 500K-entry codebook from clustered SigLIP embeddings and uses a cascaded frozen-plus-trainable codebook to unify multimodal understanding and generation with stable training and high token utilization.
-
DiSA: Diffusion Step Annealing in Autoregressive Image Generation
Autoregressive image diffusion models can use far fewer denoising steps for later tokens without losing quality, yielding 1.4-2.5x speedup from step annealing and up to 10x when combined with fewer autoregressive steps.
-
Plug-and-Play Context Feature Reuse for Efficient Masked Generation
ReCAP interleaves full model evaluations with lightweight steps that reuse cached context features, delivering up to 2.4x faster masked generation with minimal FID loss.
-
Grouping First, Attending Smartly: Training-Free Acceleration for Diffusion Transformers
GRAT accelerates pretrained diffusion transformers by grouping tokens and restricting each group's attention to neighboring blocks or criss-cross rows and columns, achieving large speedups with near-full-attention quality.
-
Marrying Autoregressive Transformer and Diffusion with Multi-Reference Autoregression
A hybrid image generator that uses an autoregressive transformer as an encoder and a diffusion model as a decoder reaches FID 1.42 on ImageNet 256x256, with an iterative multi-reference mode that improves quality over...
Discussion (0). Continue with ORCID to comment.