Token-Shuffle: Towards High-Resolution Image Generation with Autoregressive Models
akhaliq
Abstract
Autoregressive (AR) models, long dominant in language generation, are increasingly applied to image synthesis but are often considered less competitive than Diffusion-based models. A primary limitation is the substantial number of image tokens required for AR models, which constrains both training and inference efficiency, as well as image resolution. To address this, we present Token-Shuffle, a novel yet simple method that reduces the number of image tokens in Transformer. Our key insight is the dimensional redundancy of visual vocabularies in Multimodal Large Language Models (MLLMs), where low-dimensional visual codes from visual encoder are directly mapped to high-dimensional language vocabularies. Leveraging this, we consider two key operations: token-shuffle, which merges spatially local tokens along channel dimension to decrease the input token number, and token-unshuffle, which untangles the inferred tokens after Transformer blocks to restore the spatial arrangement for output. Jointly training with textual prompts, our strategy requires no additional pretrained text-encoder and enables MLLMs to support extremely high-resolution image synthesis in a unified next-token prediction way while maintaining efficient training and inference. For the first time, we push the boundary of AR text-to-image generation to a resolution of 2048x2048 with gratifying generation performance. In GenAI-benchmark, our 2.7B model achieves 0.77 overall score on hard prompts, outperforming AR models LlamaGen by 0.18 and diffusion models LDM by 0.15. Exhaustive large-scale human evaluations also demonstrate our prominent image generation ability in terms of text-alignment, visual flaw, and visual appearance. We hope that Token-Shuffle can serve as a foundational design for efficient high-resolution image generation within MLLMs.
Community
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- HiTVideo: Hierarchical Tokenizers for Enhancing Text-to-Video Generation with Autoregressive Large Language Models (2025)
- V2Flow: Unifying Visual Tokenization and Large Language Model Vocabularies for Autoregressive Image Generation (2025)
- Unleashing the Potential of Large Language Models for Text-to-Image Generation through Autoregressive Representation Alignment (2025)
- Beyond Words: Advancing Long-Text Image Generation via Multimodal Autoregressive Models (2025)
- Layton: Latent Consistency Tokenizer for 1024-pixel Image Reconstruction and Generation by 256 Tokens (2025)
- Unified Autoregressive Visual Generation and Understanding with Continuous Tokens (2025)
- GigaTok: Scaling Visual Tokenizers to 3 Billion Parameters for Autoregressive Image Generation (2025)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment:
@librarian-bot
recommend
Interesting work!
I think the "toke-shuffling" and "token-unshuffling" concepts are similar to the jigsaw puzzle solving strategy.
A few years back, we used a similar strategy of concatenating the output features from each patch along the channel dimension to prevent learning of redundant representations in the pre-training stage. However, our works [1,2] were focused on learning representations for Knee injury diagnosis from Knee MR scans in the downstream task.
References:
[1] S. Manna, S. Bhattacharya, U. Pal, "Self-supervised representation learning for detection of ACL tear injury in knee MR videos," Pattern Recognition Letters, Volume 154, 2022, Pages 37-43, https://doi.org/10.1016/j.patrec.2022.01.008.
[2] S. Manna, S. Bhattacharya and U. Pal, "Self-Supervised Representation Learning for Knee Injury Diagnosis From Magnetic Resonance Data," in IEEE Transactions on Artificial Intelligence, vol. 5, no. 4, pp. 1613-1623, April 2024, doi: 10.1109/TAI.2023.3299883.
Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper