Paper
Submissions of 2–8 pages (excluding references and supplementary materials) following ICML 2026 format, to be presented as posters with select papers invited for oral talks.
Machine Learning/CORE Unranked
ICML 2026
3rd AI for Math Workshop: Toward Self-Evolving Scientific Agents
The 3rd AI for Math Workshop (AI4Math) at ICML 2026 focuses on developing self-evolving AI agents capable of advanced mathematical reasoning and scientific discovery. It brings together researchers to explore breakthroughs in formal theorem proving, autoformalization, natural language mathematical reasoning, and multimodal scientific problem solving, with an emphasis on verifiability and reliability.
Key deadlines
Verified September 9, 2026Full paper
May 26, 2026
AoE
Workshop timeline
Submission and decisions
Full paperKey deadline
May 26, 2026 · AoE
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Challenge Submission
Competitive entries in four Codabench tracks: Semantic Alignment Evaluation, Theoretical Computer Science Proving in Lean, Visual Grounded Physics Problem Solving, and End-to-End Autoformalization and Proof Generation in Lean 4.
Research areas in scope
01
Core Themes
Formal Theorem ProvingPrecise autoformalizationAutomated mathematics in natural languageScientific problem solvingMultimodal reasoning
02
Additional Welcome Topics
Verification and measurementHuman-AI collaborationScientific agents for systems scienceCausalityFinanceBioinformaticsRelated areas
Policies worth checking twice
- Papers must be 2–8 pages (excluding references and supplementary materials)
- Submissions must follow the ICML 2026 format
- Review process is double-blind via OpenReview
- Workshop is non-archival: submissions to or accepted at other venues are welcome
- Authors may change authorship after submission (implied by camera-ready deadline and no restriction mentioned)
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026