Inference-Time Scaling for Visual AutoRegressive modeling by Searching Representative Samples
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917195930402816 |
|---|---|
| author | Tang, Weidong Wan, Xinyan Li, Siyu Wang, Xiumei |
| author_facet | Tang, Weidong Wan, Xinyan Li, Siyu Wang, Xiumei |
| contents | While inference-time scaling has significantly enhanced generative quality in large language and diffusion models, its application to vector-quantized (VQ) visual autoregressive modeling (VAR) remains unexplored. We introduce VAR-Scaling, the first general framework for inference-time scaling in VAR, addressing the critical challenge of discrete latent spaces that prohibit continuous path search. We find that VAR scales exhibit two distinct pattern types: general patterns and specific patterns, where later-stage specific patterns conditionally optimize early-stage general patterns. To overcome the discrete latent space barrier in VQ models, we map sampling spaces to quasi-continuous feature spaces via kernel density estimation (KDE), where high-density samples approximate stable, high-quality solutions. This transformation enables effective navigation of sampling distributions. We propose a density-adaptive hybrid sampling strategy: Top-k sampling focuses on high-density regions to preserve quality near distribution modes, while Random-k sampling explores low-density areas to maintain diversity and prevent premature convergence. Consequently, VAR-Scaling optimizes sample fidelity at critical scales to enhance output quality. Experiments in class-conditional and text-to-image evaluations demonstrate significant improvements in inference process. The code is available at https://github.com/WD7ang/VAR-Scaling. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_07293 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Inference-Time Scaling for Visual AutoRegressive modeling by Searching Representative Samples Tang, Weidong Wan, Xinyan Li, Siyu Wang, Xiumei Computer Vision and Pattern Recognition I.4.0; I.2.6 While inference-time scaling has significantly enhanced generative quality in large language and diffusion models, its application to vector-quantized (VQ) visual autoregressive modeling (VAR) remains unexplored. We introduce VAR-Scaling, the first general framework for inference-time scaling in VAR, addressing the critical challenge of discrete latent spaces that prohibit continuous path search. We find that VAR scales exhibit two distinct pattern types: general patterns and specific patterns, where later-stage specific patterns conditionally optimize early-stage general patterns. To overcome the discrete latent space barrier in VQ models, we map sampling spaces to quasi-continuous feature spaces via kernel density estimation (KDE), where high-density samples approximate stable, high-quality solutions. This transformation enables effective navigation of sampling distributions. We propose a density-adaptive hybrid sampling strategy: Top-k sampling focuses on high-density regions to preserve quality near distribution modes, while Random-k sampling explores low-density areas to maintain diversity and prevent premature convergence. Consequently, VAR-Scaling optimizes sample fidelity at critical scales to enhance output quality. Experiments in class-conditional and text-to-image evaluations demonstrate significant improvements in inference process. The code is available at https://github.com/WD7ang/VAR-Scaling. |
| title | Inference-Time Scaling for Visual AutoRegressive modeling by Searching Representative Samples |
| topic | Computer Vision and Pattern Recognition I.4.0; I.2.6 |
| url | https://arxiv.org/abs/2601.07293 |