![[2602.14041] BitDance: Scaling Autoregressive Generative Models with Binary Tokens](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.14041] BitDance: Scaling Autoregressive Generative Models with Binary Tokens
Summary
BitDance introduces a novel autoregressive image generator that utilizes binary tokens for enhanced efficiency and performance in generating high-resolution images.
Why It Matters
This research is significant as it addresses the challenges of scaling autoregressive generative models, offering a more efficient method for image generation. By employing binary tokens and innovative decoding techniques, BitDance achieves superior performance with fewer parameters, making it a valuable contribution to the field of generative AI.
Key Takeaways
- BitDance uses binary tokens to represent a vast state space, enhancing model expressiveness.
- The new next-patch diffusion method allows for parallel token prediction, significantly speeding up inference.
- BitDance achieves state-of-the-art performance on ImageNet with fewer parameters and faster generation times.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.14041 (cs) [Submitted on 15 Feb 2026] Title:BitDance: Scaling Autoregressive Generative Models with Binary Tokens Authors:Yuang Ai, Jiaming Han, Shaobin Zhuang, Weijia Mao, Xuefeng Hu, Ziyan Yang, Zhenheng Yang, Huaibo Huang, Xiangyu Yue, Hao Chen View a PDF of the paper titled BitDance: Scaling Autoregressive Generative Models with Binary Tokens, by Yuang Ai and 9 other authors View PDF Abstract:We present BitDance, a scalable autoregressive (AR) image generator that predicts binary visual tokens instead of codebook indices. With high-entropy binary latents, BitDance lets each token represent up to $2^{256}$ states, yielding a compact yet highly expressive discrete representation. Sampling from such a huge token space is difficult with standard classification. To resolve this, BitDance uses a binary diffusion head: instead of predicting an index with softmax, it employs continuous-space diffusion to generate the binary tokens. Furthermore, we propose next-patch diffusion, a new decoding method that predicts multiple tokens in parallel with high accuracy, greatly speeding up inference. On ImageNet 256x256, BitDance achieves an FID of 1.24, the best among AR models. With next-patch diffusion, BitDance beats state-of-the-art parallel AR models that use 1.4B parameters, while using 5.4x fewer parameters (260M) and achieving 8.7x speedup. For text-to-image generation, BitDance trains on large-scale multimoda...
