Full papers
Up to 9 pages, excluding references and an optional appendix; submissions may present new results, works in progress, negative results, empirical evaluations, or forward-looking position papers.
Non-Autoregressive Language Models for Fast and Flexible Text Generation
NonAR-LM 2026 is a workshop co-located with COLM 2026 in San Francisco, focusing on non-autoregressive language models for fast and flexible text generation. It explores alternatives to left-to-right next-token prediction, including diffusion, flow matching, and any-order autoregression, with an emphasis on parallel decoding, controllability, and global coherence in sequence generation.

Full paper
July 2, 2026
AoE
Conference timeline
Full paperKey deadline
July 2, 2026 · AoE
Paper fit
A strong submission should clearly identify its contribution and evaluate it appropriately.
Up to 9 pages, excluding references and an optional appendix; submissions may present new results, works in progress, negative results, empirical evaluations, or forward-looking position papers.
Shorter submissions are equally welcome.
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026