CFP: Trust4RSI-Agent: AAAI 2027 Workshop on Trustworthy Evolution of AI Agents
We are delighted to invite you to submit your work to Trust4RSI-Agent, held in conjunction with AAAI 2027 in Montréal, Canada. The workshop brings together researchers in agent learning, ML systems, and AI safety to explore how AI agents can evolve their memories, tools, skills, and policies while remaining trustworthy.
*TOPICS OF INTEREST*
- Trustworthy memory evolution: provenance, consolidation, forgetting, and prevention of persistent errors. - Trustworthy tool and skill evolution: access control, sandboxing, verification, and security. - Trustworthy model and policy evolution: safe updates, regression testing, rollback, and reliable self-critique. - Evaluation of evolving agents: long-horizon benchmarks, monitoring, auditing, and human oversight.
We welcome methods, system designs, benchmarks, negative results, and position papers.
*SPEAKERS AND PANELISTS*
- Yoshua Bengio (Université de Montréal, Mila, LawZero); - Mengdi Wang (Princeton); - Yilun Du (Harvard); - Hugo Larochelle (Mila); - Zheqing (Bill) Zhu (Pokee AI); - Yue Zhao (USC); - Ying Liu (Scale AI).
*ORGANIZERS*
- Haolun Wu (Stanford & Mila); - Negar Arabzadeh (UC Berkeley); - Wenyue Hua (Microsoft Research); - Bang Liu (Mila & Université de Montréal); - Laurent Charlin (Mila & HEC Montréal).
*KEY DATES*
- Submission deadline: November 20, 2026 (AoE) - Author notification: December 2, 2026 - Workshop: One day during February 22–23, 2027; exact day TBA
*SUBMISSION DETAILS*
Full papers (8 pages) and short or position papers (4 pages) are welcome. Submissions follow a double-blind review process, and the workshop is non-archival.
Submit your paper via OpenReview. https://openreview.net/group?id=AAAI.org/2027/Workshop/Trust4RSI-Agent
Full call, submission instructions, and program details: https://trust4rsi-agent.github.io/AAAI2027/
Limited workshop registration reimbursement grants are available for student authors of accepted papers.
Please share this call with interested colleagues and students. We look forward to your submissions!
Best, Negar Arabzadeh
participants (1)
-
Negar Arabzadehghahyazi