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.
NLP/CORE Unranked
NonAR-LM 2026
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.
Key deadlines
Verified September 9, 2026Full paper
July 2, 2026
AoE
Conference timeline
Submission and decisions
Full paperKey deadline
July 2, 2026 · AoE
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Shorter submissions
Shorter submissions are equally welcome, with the same formatting requirements and page limit flexibility.
Research areas in scope
01
Modeling & Training
New model classes and training objectives — discrete diffusion, uniform-state diffusion, flow-based, and any-order approaches.
02
Inference & Sampling
Inference-time algorithms: iterative refinement, parallel decoding, controllable and constrained generation, planning and correction.
03
Evaluation & Efficiency
Evaluation beyond left-to-right likelihood, plus parallel generation, latency-constrained inference, and systems for scaling.
04
Applications
Applications across language, code, and biological sequences — including comparative studies of when iterative models help.
Policies worth checking twice
- Submissions are non-archival and do not preclude publishing elsewhere.
- Double-blind review: each submission receives at least three reviews.
- Use of COLM 2026 or NeurIPS 2026 LaTeX template in submission mode is required.
- Submissions are handled via OpenReview.
- Six spotlight talks are selected from submissions; all accepted work is presented as posters.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026