mlsanityData to 14 Oct 2026

Balancing Image Compression and Generation with Bootstrapped Tokenization

arXiv 2606.05552 ↗First posted 4 Jun 2026cs.LG cs.AI cs.GRcited by 1

Not on the map. Why?

Abstract (arXiv)

Despite progress in image tokenization, standard methods encode redundant information by mixing all granularities within each token, thus redundancy persists between tokens. The mix of information of different granularity also complicates the training of generators. This paper introduces SelfBootTok, a method that resolves this by cleanly decomposing information into global and local token groups. Through self-bootstrapped learning, the model predicts local details exclusively from global tokens, shifting the burden of visual details from the generator to the tokenizer. Consequently, our generator is far more efficient, requiring only global tokens and reducing computation by approximately 40%, while delivering superior reconstruction and generation. Moreover, this paradigm scales elegantly: by leveraging more data or parameters to self-supervise local representation learning, SelfBootTok achieves a new state-of-the-art gFID score of 1.56 using only 64 tokens.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 62 of its 66 references that are in the map. First posted on arXiv 4 Jun 2026.

Builds on

62 of 66 references.

References · 62

  1. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  2. 2023GPT-4 Technical Reportcited by 18,396
  3. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  4. 2023Visual Instruction Tuningcited by 7,869
  5. 2023BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Modelscited by 6,137
  6. 2022Scalable Diffusion Models with Transformerscited by 6,054
  7. 2018A Style-Based Generator Architecture for Generative Adversarial Networkscited by 6,013
  8. 2022Flow Matching for Generative Modelingcited by 5,702
  9. 2017Neural Discrete Representation Learningcited by 4,712
  10. 2022Flamingo: a Visual Language Model for Few-Shot Learningcited by 4,386
  11. 2019Analyzing and Improving the Image Quality of StyleGANcited by 3,588
  12. 2020Taming Transformers for High-Resolution Image Synthesiscited by 3,253
38 more
  1. 2023InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuningcited by 2,876
  2. 2023Video-LLaVA: Learning United Visual Representation by Alignment Before Projectioncited by 1,402
  3. 2025SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Featurescited by 1,188
  4. 2022MaskGIT: Masked Generative Image Transformercited by 1,099
  5. 2023Video-ChatGPT: Towards Detailed Video Understanding via Large Vision and Language Modelscited by 1,062
  6. 2023Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecturecited by 1,039
  7. 2024Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Predictioncited by 1,023
  8. 2023VideoChat: Chat-Centric Video Understandingcited by 857
  9. 2024Autoregressive Model Beats Diffusion: Llama for Scalable Image Generationcited by 777
  10. 2024SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformerscited by 737
  11. 2022Autoregressive Image Generation using Residual Quantizationcited by 722
  12. 2021Vector-quantized Image Modeling with Improved VQGANcited by 658
  13. 2024Autoregressive Image Generation without Vector Quantizationcited by 651
  14. 2023Language Model Beats Diffusion -- Tokenizer is Key to Visual Generationcited by 640
  15. 2024LLaVA-NeXT-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal Modelscited by 584
  16. 2023OpenFlamingo: An Open-Source Framework for Training Large Autoregressive Vision-Language Modelscited by 565
  17. 2022All are Worth Words: A ViT Backbone for Diffusion Modelscited by 502
  18. 2025Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Modelscited by 393
  19. 2024InternVideo2: Scaling Foundation Models for Multimodal Video Understandingcited by 317
  20. 2024Discrete Flow Matchingcited by 317
  21. 2024An Image is Worth 32 Tokens for Reconstruction and Generationcited by 302
  22. 2022MAGE: MAsked Generative Encoder to Unify Representation Learning and Image Synthesiscited by 227
  23. 2023SVIT: Scaling up Visual Instruction Tuningcited by 162
  24. 2023Flow Matching in Latent Spacecited by 160
  25. 2024AuroraCap: Efficient, Performant Video Detailed Captioning and a New Benchmarkcited by 140
  26. 2019Consistency Regularization for Generative Adversarial Networkscited by 136
  27. 2025FlexTok: Resampling Images into 1D Token Sequences of Flexible Lengthcited by 110
  28. 2021Regularizing Generative Adversarial Networks under Limited Datacited by 107
  29. 2024LongLLaVA: Scaling Multi-modal LLMs to 1000 Images Efficiently via a Hybrid Architecturecited by 102
  30. 2025Representation Entanglement for Generation: Training Diffusion Transformers Is Much Easier Than You Thinkcited by 102
  31. 2024MaskBit: Embedding-free Image Generation via Bit Tokenscited by 98
  32. 2025Masked Autoencoders Are Effective Tokenizers for Diffusion Modelscited by 96
  33. 2024ImageFolder: Autoregressive Image Generation with Folded Tokenscited by 94
  34. 2024FlowAR: Scale-wise Autoregressive Image Generation Meets Flow Matchingcited by 80
  35. 2025GigaTok: Scaling Visual Tokenizers to 3 Billion Parameters for Autoregressive Image Generationcited by 57
  36. 2024CuMo: Scaling Multimodal LLM with Co-Upcycled Mixture-of-Expertscited by 56
  37. 2024SoftVQ-VAE: Efficient 1-Dimensional Continuous Tokenizercited by 56
  38. 2025Democratizing Text-to-Image Masked Generative Models with Compact Text-Aware One-Dimensional Tokenscited by 50

50 most cited of 62 All 62 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026MixTok: Adaptive Visual Tokenization with Global 1D Tokens and Sparse 2D Refinementcited by 0