Original Research Track
4–8 page papers describing new algorithms, systems, or scientific findings enabled by AI, with unlimited references and appendices.
Machine Learning/CORE Unranked
ICML2026-AI4Science 2026
ICML 2026 AI for Science Workshop
The ICML 2026 workshop 'AI Scientists – Tools, Co-authors, or Founders?' explores the evolving role of AI in scientific discovery, examining systems that range from passive tools to autonomous agents capable of planning experiments and drafting papers. It brings together researchers from machine learning and domain sciences to establish shared definitions, evaluation criteria, and governance principles for human–AI collaboration in science.
Key deadlines
Verified September 9, 2026Abstract registration
May 7, 2026
AoE
Full paper
May 8, 2026
AoE
Workshop timeline
Submission and decisions
Abstract registrationKey deadline
May 7, 2026 · AoE
Full paperKey deadline
May 8, 2026 · AoE
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Attention Track
4–8 page position papers articulating opportunities, risks, and critical perspectives on AI-driven science.
Highlight Track
4–8 page surveys, benchmarks, or synthesis papers consolidating progress on specific AI4Science topics.
Education Track
4–8 page tutorials, curricula, notebooks, or review papers providing open educational resources for building AI scientists.
Dataset Proposal Competition
2-page proposals for new datasets capturing the full scientific stack from planning to robotic execution.
AI Scientist Competition
2-page proposals for AI scientist systems, sponsored by Samsung Advanced Institute of Technology (SAIT).
Research areas in scope
01
Original Research Track
Closed-loop design-build-test workflows and autonomous labsMultiscale modeling of molecules, materials, or climate systemsGenerative models for catalysts, battery materials, or biological sequencesScientific reasoning agents that incorporate symbolic, causal, or geometric priorsNew benchmarks or datasets for evaluating AI scientists
02
Attention Track
Policies for credit assignment and authorship with AI collaboratorsPractical guidance for responsible deployment in physical labsOpportunities at the intersection of disciplines that rarely meetFuture research agendas for sociotechnical governance
03
Highlight Track
Benchmark suites for cooperative human–AI scientific teamsLessons from AI-assisted drug discovery campaignsComparative studies of autonomous experimentation platformsEvaluations of agent-based scientific reasoning
04
Education Track
Tutorials, curricula, notebooks, or review papers for building AI scientistsOpen educational resources to lower barriers for students and practitioners
Policies worth checking twice
- Submissions must be double-blind; identifying information must be removed from main papers and supplementary files.
- All submissions use the official ICML 2026 style files with double-blind option.
- Submissions may include work in progress, previously unpublished results, or recently published science-journal articles.
- The workshop is nonarchival; accepted papers remain eligible for future archival venues.
- Dual submission to multiple ICML workshops is not permitted.
- Appendices may include implementation details but are not required to be read by reviewers.
- Submissions must update the ICML template footnote to: 'Submitted to/Accepted at/Published in the AI for Science workshop (ICML 2026).'
- Reviews are conducted by 300+ reviewers and 50+ area chairs, ensuring at least 2–3 expert evaluations per paper.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026