About the Workshop
From AI scientists to autonomous research
Scientific discovery is entering a phase of profound transformation. Advances in foundation models, agentic AI, and embodied robotics now make it feasible to automate substantial portions of the research lifecycle: literature synthesis, hypothesis generation, experimental design, physical experimentation, data analysis, manuscript drafting, and even peer review.
Unlike adjacent venues that address a single stage of discovery or a single scientific domain, AutoResearch takes the full research lifecycle as its unit of analysis, and pairs autonomous AI Scientists with embodied Robot Scientists that can act in physical laboratories. For the first time at a NeurIPS workshop, we add a submission track for AI-authored research and an AI peer-review pilot, building on the AI Scientists Conference (AISC 2026) protocol and aligned with the NeurIPS 2026 AI-Assisted Reviewing Experiment.
AI + Robot Scientists
Agentic reasoning and physical experimentation in one venue, not virtual-only or robotics-only.
Full Research Lifecycle
From idea, to experiment, to manuscript, to review, treated as one integrated system.
AI-Authored Track + AI Review
A dedicated track for AI-led papers and a paired, public AI peer-review pilot.
Why now?
Three developments in the past 18 months make this workshop urgent:
- The first fully AI-generated manuscript passing peer review at an ICLR 2025 workshop (Sakana AI Scientist v2).
- The launch of the AI Scientists Conference (AISC 2026), the first venue where the entire reviewing pipeline is run by AI agents.
- The NeurIPS 2026 AI-Assisted Reviewing Experiment, which formalizes the study of LLM-augmented peer review at the flagship venue itself.
Scope
Topics of interest
We welcome submissions on, but not limited to:
Call for Papers
Two tracks, one autonomous future
All submissions use OpenReview and the official NeurIPS 2026 LaTeX template. Accepted papers are non-archival and presented as posters; selected papers receive spotlight talks.
Human-led / Human-AI research
- Tiny papers up to 2 pages: perspectives, problem statements.
- Short papers up to 4 pages: work in progress, technical reports.
- Full papers 4 to 8 pages: original research.
References and appendices excluded from page limits.
AI-authored / AI-led research
Research primarily produced by AI agents, with or without human assistance. Submissions must:
- Declare the agentic pipeline used.
- Provide a reproducible transcript (logs, code, prompts).
- Disclose human involvement and level of supervision.
AI Review Pilot
Alongside human review, submissions may opt in to receive AI-generated reviews in addition to (never in place of) human reviews. AI reviews are released publicly with the human reviews and outcomes, producing a paired benchmark for AI peer-review research. Decisions rest with human program-committee chairs. A Best AI Review prize will be awarded.
Submission
OpenReview (link TBA after acceptance). NeurIPS 2026 LaTeX template. Non-archival; dual submission allowed where permitted.
Awards
3 spotlight talks across tracks, 2 Best Paper Awards (one per track), and 1 Best AI Review award, with cash prizes from sponsors.
Timeline
Important dates
All dates are tentative and will be finalized to align with the NeurIPS 2026 calendar (Anywhere on Earth).
- Jul 2026Workshop acceptance & website launch; first call for papers
- Aug 29, 2026Paper submission deadline (tentative)
- Sep 2026Review period (human + AI pilot)
- Sep 29, 2026Notification of acceptance
- Dec 2026Workshop day at NeurIPS 2026
Program
Tentative schedule
A one-day program of keynotes, contributed talks, two poster sessions, and two panels. Final schedule TBA.
| Time | Session |
|---|---|
| 09:00–09:15 | Opening remarks |
| 09:15–10:15 | Keynotes: AI Scientists & Robot Scientists |
| 10:15–10:45 | Track A spotlight talks |
| 10:45–11:15 | Coffee & Poster session I |
| 11:15–12:15 | Keynotes: Foundation Models & Domain Science |
| 12:15–14:00 | Track B briefing & lunch |
| 14:00–15:00 | Keynote & Track B spotlight talks |
| 15:00–15:30 | Coffee & Poster session II + AI-review board |
| 15:30–17:30 | Panel 1: AI → Robot Scientists · Panel 2: AI-Authored Papers & AI Reviewers |
| 17:30–18:00 | Awards & closing remarks |
Invited Speakers and Panelists
Speakers and Panelists
Confirmed speakers and panelists for AutoResearch 2026.
Prof. James Zou
Stanford, USA
Prof. Le Cong
Stanford, USA
Prof. Markus J. Buehler
MIT, USA
Prof. Graham Neubig
CMU, USA
Prof. Ross King
Cambridge, UK
Dr. David Ha
Sakana AI, Japan
Dr. Vivek Natarajan
Google DeepMind, USA
Dr. Cong Lu
Google DeepMind, USA
Prof. Pulkit Agrawal
MIT, USA
Prof. Lili Pan
UESTC, China
Dr. Bodhisattwa Majumder
Allen Institute for AI, USA
Dr. Yuan Cao
Unreasonable Labs, USA
Dr. Hirotaka Sugiura
University of Tokyo, Japan
Ms. Kexin Huang
Stanford, USA
Mr. Bowen Gao
Tsinghua University AIR, China
Prof. Jure Leskovec
Stanford, USA (tentative)
Team
Organizers
Organizing committee for AutoResearch 2026.
Guowei Huang
The University of Manchester, UK
Xiang Hu
China Life R&D Center & Westlake University, China
Haokun Liu
University of Chicago, USA
Shiyi Du
Carnegie Mellon University, USA
Liang Yan
University of Washington, USA
Yanlan Kang
Fudan University, China
Pengsong Zhang
University of Toronto, Canada
Program Committee
Confirmed members, with more invitations in progress.
Heng Zhang
Istituto Italiano di Tecnologia, Italy
Pengsong Zhang
University of Toronto, Canada
Jiayuan Liu
Carnegie Mellon University, USA
Xinge Liu
University of Toronto, Canada
Ang (Leo) Li
University of Toronto, Canada
Lili Pan
UESTC, China
Lichao Sun
Lehigh University, USA
More
More to be invited
Advisory Board
To be announced.
To be announced
To be announced
Support
Sponsors
Sponsorship supports registration grants for under-represented students, Best Paper and Best AI Review prizes, and AI-review-pilot compute. Interested in sponsoring? Get in touch.
Get in touch
Contact
For any questions, please reach out to the organizing committee. A dedicated workshop contact will be posted here after acceptance.
Contact: to be announced
Participants are expected to abide by the NeurIPS Code of Conduct.