Short papers
Up to 4 pages, not including references or appendix.
NLP/CORE Unranked
ACL 2026
Second Workshop on Customizable NLP (CustomNLP4U)
The Second Workshop on Customizable NLP (CustomNLP4U) @ ACL 2026 focuses on developing language models that can be tailored to individual users, groups, or organizations, addressing gaps in generalist models across domains like law, healthcare, and finance. It explores methodological, ethical, and societal aspects of customization, including privacy, bias, data ownership, and open science practices.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Long papers
Up to 8 pages, not including references or appendix.
Research areas in scope
01
Topics of Interest
Data: Data collection, processing, and annotation to aid model customization; analysis of public and private datasets to guide the development of customized models; privacy, and copyright.Modeling: New pretraining, fine-tuning, alignment, continuous learning, inference methods for customizing language models; privacy and efficiency for customization. New modeling paradigms for customization such as model ensembles, federated learning, nonparametric models, etc.; customizing models with agentic systems, test-time compute strategies, etc.Evaluation of generalist, non-customized models, identifying their shortcomings for varied use-cases; evaluation of customized models.Applications: e.g. information seeking, writing assistance, AI assistants etc.; applications in sensitive domains such as law, education, healthcare, finance, etc.; models for communities reflecting sociolects, dialects or other language varieties.Broader Impacts: privacy, security, and copyright; intrusiveness, unintended biases; invisibility versus hypervisibility due to customization.Open Science: Best practices for open and reproducible science of customizable LMs: dataset release and licensing, open-sourcing models, related privacy, copyright and policy issues.
Policies worth checking twice
- Paper submission must be anonymized and uploaded as a single PDF to OpenReview.
- Authors may not modify the provided LaTeX style files or use templates from other conferences.
- Posting papers on preprint servers like ArXiv is permitted.
- The review process is double-blind.
- Authors may submit papers under review at other venues (double submissions), but must email the organizers to declare this.
- Submissions can be archival (indexed in proceedings) or non-archival (listed only on workshop website); authors must notify organizers by May 27, 2026, to choose archival status.
- Non-archival papers will not be included in the ACL Anthology and their PDFs will not be made public.
- Reviewers are not required to read the appendix, though unlimited pages are allowed for references and appendix.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026