Program
(All Times are Banff time)
Opening Remarks 9:15-9:30
9:30
Nolan Bard
Title: Reflections on The Hanabi Challenge
The Hanabi Challenge first appeared five years ago in February 2019. Since then, AI practitioners have explored the challenge, yielding insights in cooperative multi-agent learning, ad-hoc teamwork, and emergent communication. This talk looks back on the challenge, what has been learned from it so far, and how those learnings can inform practitioners going forward.
Coffee Break 10:30-11:00
11:00
Paper Presentations
Osher Elhadad, Matan Shamir and Reuth Mirsky / EDGE: Ensemble Methods for Dynamic Goal Recognition
Caroline Wang, Arrasy Rahman, Ishan Durugkar, Elad Liebman and Peter Stone / Towards Bridging Cooperative MARL and Ad Hoc Teamwork
Dylan Cope and Peter Mcburney / Learning Translations: Emergent Communication Pretraining for Cooperative Language Acquisition
Shibei Zhu, Tran Nguyen Le, Samuel Kaski and Ville Kyrki. Online Learning of Human Constraints from Feedback in Shared Autonomy
Muhammad Arrasy Rahman, Jiaxun Cui and Peter Stone. Minimum Coverage Sets for Training Robust Ad Hoc Teamwork Agents
Lunch break 12:15-13:15
13:15
Hanabi Session and Discussion
15:00
Paper Presentations
Sebastiaan De Peuter and Samuel Kaski / Helping Designers Solve Design Problem with Ad Hoc Collaborative Assistants
João G. Ribeiro, Gonçalo Rodrigues, Alberto Sardinha and Francisco S. Melo / TEAMSTER: Model-based reinforcement learning for ad hoc teamwork
Muhammad Arrasy Rahman, Ignacio Carlucho, Niklas Hoepner and Stefano Albrecht / Graph-Based Policy Learning in Partially Observable Open Ad Hoc Teamwork
Coffee Break 15:45-16:15
Concluding Remarks 16:15-16:30