Taming the Entropy Cliff: Variable Codebook Size Quantization for Autoregressive Visual Generation
cs.CV, cs.AI, cs.LG
Submitted: 2026-05-07
Updated: 2026-11-05
Terminology
Sources
- LLaMA: Open and Efficient Foundation Language Models
- ResTok: Learning Hierarchical Residuals in 1D Visual Tokenizers for Autoregressive Image Generation
- SpectralAR: Spectral Autoregressive Visual Generation
- DetailFlow: 1D Coarse-to-Fine Autoregressive Image Generation via Next-Detail Prediction
- Randomized Autoregressive Visual Generation
- Open-MAGVIT2: An Open-Source Project Toward Democratizing Auto-regressive Visual Generation
- Scalable Image Tokenization with Index Backpropagation Quantization
- FlexTok: Resampling Images into 1D Token Sequences of Flexible Length
- One-D-Piece: Image Tokenizer Meets Quality-Controllable Compression
- Selftok: Discrete Visual Tokens of Autoregression, by Diffusion, and for Reasoning
- Classifier-Free Diffusion Guidance
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs