Full papers
Up to 5 pages of main text (references and appendices excluded); accepted papers may expand to 6 pages in camera-ready version.
Machine Learning/CORE Unranked
Interp4Discovery 2026
NeurIPS 2026 Workshop on Interpretability for Discovery
Interp4Discovery 2026 is a NeurIPS workshop focused on using model interpretability as a tool for scientific discovery, bringing together interpretability researchers and domain scientists to transform learned AI representations into novel, testable knowledge. The workshop welcomes work that uncovers new insights in fields such as biology, climate science, astronomy, and medicine through interpretability methods.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Fast-track submissions
Papers already accepted at NeurIPS 2026 may be submitted for fast-track consideration, subject to additional review for workshop fit.
Research areas in scope
01
Research Scope
Methodological approaches to interpretability for discoveryEmpirical studies using interpretability to uncover new knowledgeTheoretical frameworks for interpretability-driven discoveryPosition papers on interpretability as a scientific methodApplications in protein structure prediction, clinical forecasting, astronomy, climate science, animal communication, and strategic games
Policies worth checking twice
- Submissions must be double-blind: remove author names, affiliations, acknowledgments, and identifying details; cite prior work in third person.
- Submissions must include a responsible-use statement covering societal impacts and mitigations; missing statements lead to desk rejection.
- Work previously presented at non-archival venues (including workshops) is allowed; work accepted at archival venues is not allowed, except for NeurIPS 2026 fast-track submissions.
- Submissions under review at other venues (e.g., ICLR, NeurIPS) are permitted.
- Accepted papers may use one additional main-text page (up to 6 pages total) in the camera-ready version to incorporate reviewer feedback.
- Authors are strongly encouraged to open source code, data, models, and demos; anonymous hosting via anonymous.4open.science or anonymous Hugging Face accounts is recommended.
- OpenReview accounts require institutional email; non-institutional accounts may take up to two weeks to approve.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Official workshop siteCFP03 · Schedule Full-day program The complete timetable and program notes in a foc
Last verified September 9, 2026