Journals and conferences / International Conference on Learning Representations (ICLR)
International Conference on Learning Representations (ICLR) AI disclosure statement
ICLR requires authors to say how they used LLMs both in the submission form and in a mandatory AI disclosure section in the paper, using a list of research subtasks that are either required or recommended to disclose. Authors and reviewers stay fully responsible for their content, and LLM-caused falsehoods, plagiarism or hallucinated references can lead to desk rejection.
Checked against International Conference on Learning Representations (ICLR)'s own pages on 9 October 2026 · See the full rules
Write yours
International Conference on Learning Representations (ICLR) doesn't publish a fixed template, so this uses a general format that covers what it asks for.
International Conference on Learning Representations (ICLR)'s own wording
In this work we used generative AI for [required-disclosure tasks]; we did not use it for [other required tasks], and [remaining required tasks] do not apply. We also used it for [recommended-disclosure tasks]. We reviewed all AI-assisted work [describe checks, e.g. manual plagiarism survey of LLM ideas, code tested by two authors] and take responsibility for all content, including text, claims or artifacts made with generative AI. (The ICLR paper template includes this boilerplate; exact format is optional.)
Summarised from International Conference on Learning Representations (ICLR)'s guidance.
The rules at a glance
- When you need one
- Any use of an LLM must be disclosed. Disclosure is required for tasks such as generating synthetic data, developing theory, formulating or proving mathematical claims, proposing hypotheses, designing experiments, implementing methods, translation, cleaning datasets, qualitative analysis and interpreting results; it is recommended for tasks such as drafting or editing text, code, figures, literature summaries, brainstorming and formatting references. Reviewers and area chairs must also disclose LLM use in reviews.
- Framework
- ICLR list of research subtasks split into 'Required' and 'Recommended' disclosure
- Where it goes
- A mandatory AI disclosure section in the paper (not counted toward the page limit), plus the paper submission form.
- Must include
- which tasks (from the required list) generative AI was used for
- which required-list tasks AI was not used for or that are not applicable
- any recommended-list tasks AI was used for
- how AI-assisted work was reviewed/verified
- statement that authors take responsibility for the final content
- Also note
- authors are responsible for all content; LLM-produced falsehoods, plagiarism or misrepresentation are Code of Ethics violations and may cause desk rejection
- a human author is required even if an LLM produced the entire research
- papers with confirmed hallucinated references were desk rejected (ICLR 2026)
- reviewers must not use LLMs in ways that breach confidentiality of submissions
- hidden prompt injections in papers aimed at LLM reviewers are treated as collusion
- ICLR ran LLM-content detectors on submissions and reviews to flag cases for area chairs
If you didn't use AI
Effectively yes for required tasks: the example disclosure states which required-disclosure tasks AI was not used for or are not applicable.
Sources
- https://iclr.cc/Conferences/2027/AIPolicyForAuthors
- https://blog.iclr.cc/2025/08/26/policies-on-large-language-model-usage-at-iclr-2026/
Your unit outline or the journal's author guide wins if it says something different. Spotted a change? Tell us.