titleNext-Concept Modeling: Fast, Parallel Coarse-to-Fine Text Generation
authorsHermann Kumbong, Ishan S. Khare, Iddah Mlauzi, Andrew Woen, Charles Ding, E. Kelly Buchanan, Simran Arora, Azalia Mirhoseini, Christopher Ré
date2026-10-09
venuearXiv (preprint, not peer reviewed)
idarXiv:2610.14162
affiliationStanford University
upvotes19 (Hugging Face Daily Papers, 2026-11-01)
linksabs pdf hf code
abstractDiffusion language models (DLMs) offer parallel sampling for faster generation than autoregressive models, an increasingly important advantage as inference becomes a dominant computational workload. However, DLMs degrade quickly with fewer sampling steps, process the full sequence at every step, and require many passes over the dataset to reach good quality. We introduce Next-Concept Modeling (NCM), a two-stage, coarse-to-fine framework: a multiscale residual VQ-VAE encodes text into a hierarchy of discrete codes, and a transformer generates one level at a time, predicting all codes within a level in parallel. At matched parameter scales, NCM achieves better generative perplexity than the strongest DLM baseline on LM1B while running 21times faster. On OpenWebText, NCM is competitive with the strongest DLM while running 10times faster; on XSum, it matches the strongest DLM's ROUGE-L while being 31times faster. On the reasoning task of Sudoku, NCM outperforms autoregressive and diffusion baselines across all difficulty levels, using fewer function evaluations. On the reasoning task of Sudoku, NCM outperforms autoregressive and diffusion baselines across all difficulty levels, using fewer function evaluations.

cite

@misc{kumbong2026next,
  title={Next-Concept Modeling: Fast, Parallel Coarse-to-Fine Text Generation},
  author={Hermann Kumbong and Ishan S. Khare and Iddah Mlauzi and Andrew Woen and Charles Ding and E. Kelly Buchanan and Simran Arora and Azalia Mirhoseini and Christopher Ré},
  year={2026},
  eprint={2610.14162},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2610.14162}
}