AI Scientist Competition
AI has shifted from passive assistant to active agent: systems now generate hypotheses, design experiments, orchestrate robots, and draft papers. The AI Scientist Competition focuses exclusively on these autonomous or semi-autonomous scientific systems. (If you are proposing a dataset, please visit the Dataset Competition page.)
Winning AI Scientist proposals share $10K in prizes sponsored by Xaira Therapeutics, present during the workshop, and help shape emerging benchmarks and governance for trustworthy AI-driven discovery.
What to Include
- Scope & autonomy: Explain which stages of the scientific workflow (ideation, planning, simulation, execution, analysis, writing) your system covers and how decisions are made.
- Architecture & tooling: Describe model components, agent orchestration, retrieval or tool use, robotic integrations, and human oversight checkpoints.
- Evaluation roadmap: Provide benchmarks, metrics, ablation plans, or verification procedures that demonstrate novelty, robustness, and trustworthiness.
- Governance & safeguards: Detail how you manage safety, privacy, biosafety, or dual-use risks, and how attribution/credit is recorded for human collaborators.
Submission Guidelines
- Length: 2 pages of main text (PDF) with unlimited references/appendices.
- Format: ICML 2026 style file (double-blind option). Update the template footnote to “Submitted to the AI for Science workshop (ICML 2026).”
- Anonymity: Fully double blind. Remove identifiable metadata, URLs, or repositories.
- Submission Portal: OpenReview → AI Scientist Competition.
- Review: At least two expert reviewers plus an area chair; top proposals earn spotlight talks and prize consideration.
Evaluation Criteria
- Scientific ambition & coverage of the end-to-end discovery pipeline.
- Technical feasibility of algorithms, agents, and hardware/software integrations.
- Evaluation rigor (benchmarks, metrics, reproducibility).
- Safety, governance, and transparency (documentation, provenance, alignment with policies).
- Potential impact on real labs or cross-domain collaborations.
Timeline (Anywhere on Earth)
- Abstract deadline: April 21, 2026
- Submission deadline: April 24, 2026
- Notification: May 15, 2026
- Camera-ready / spotlight materials: May 29, 2026
- Live spotlights: Workshop day at ICML 2026 (July, exact date TBA)
Questions? Email ai4sciencecommunity@gmail.com. If you need introductions to collaborators (e.g., robotics labs or evaluation partners), reach out early so we can help.