This is the third edition of the Bayes-duality workshop focusing on the development of AI that learns adaptively, robustly and continuously, like humans. The main goal of the meeting is two folds. First, to showcase our research conducted at RIKEN-AIP (first phase ends in 2025), and, second, to discuss (and refine) the long-term goals of the JST-CREAST project and foster collaborations among the attendees. Only Part 2 of the workshop is open to public (through live streaming) and will feature invited talks by world-leading researchers working on related topics. Part 1 and 3 will be closed meetings exclusively for the collaborators of the Bayes-duality project.

• Part 1 (June 23-24) Discussion sessions on special topics at RIKEN AIP Nihonbashi
• Part 2 (June 25-27) Invited Talks at RIKEN AIP Nihonbashi (streamed live on Zoom, register here)
• Part 3 (June 30-July 4) CREST-ANR closed meeting at LecTore Atami Koarachi in Atami.

Similarly to the previous edition, we hope that the proposed activities help attendees increase their research's impact, outreach and help them network. This workshop is supported by RIKEN-AIP subsidy budget and the JST CREST Grant JPMJCR2112 on “A new Bayes-duality principle for adaptive, robust, and lifelong learning of AI systems”.

All talks in Part 2 (June 25-27) will be streamed live and you can register at the Doorkeeper. In-person attendance is limited to RIKEN-AIP researchers and invited guests. Videos will also be made publicly available, subject to the consent of speakers, attendees, and organizers.

Participants

Invited Speakers (in alphabetical order)

ABI Team, RIKEN

• Emtiyaz Khan, RIKEN
• Thomas Möllenhoff, RIKEN
• Hugo Monzón, RIKEN
• Keigo Nishida, RIKEN
• Christopher Anders, RIKEN
• Yohan Jung, RIKEN

Math-Science Team, RIKEN

• Kenichi Bannai, Keio University/RIKEN
• Eren Mehmet Kıral, Keio University
• Koichi Tojo, RIKEN
• Kei Hagihara, Keio University

HPC Team, Institute of Science Tokyo

• Rio Yokota, Institute of Science Tokyo
• Cong Bai, Institute of Science Tokyo
• Hiro Ishii, Institute of Science Tokyo
• Eiki Shimizu, The Institute of Statistical Mathematics

Stat-Theory Team, INRIA Grenoble Rhône-Alpes, France

• Julyan Arbel, INRIA, Grenoble Rhone-Alpes

Attendees

• Martin Mundt, TU Darmstadt, Germany
• Siddharth Swaroop, Harvard University, USA
• Pierre Alquier, ESSEC Business School, Singapore
• Nico Daheim, TU Darmstadt, Germany
• Darmesh Tailor, University of Amsterdam, Netherlands
• Subarnaduti Paul, University of Bremen, Germany
• Jonghyun Choi, Seoul National University, South Korea
• Kai Arulkumaran, Araya Inc., Japan
• Frank Nielsen, Sony CSL, Japan
• Joe Austerweil, University of Wisconsin-Madison, USA
• Sarath Chandar, Polytechnique Montreal, Canada
• Weiwei Pan, Harvard University, USA

Program

[The detailed program]

	June 23		June 24		June 25		June 26		June 27
09:50-10:00									Masashi Sugiyama
10:00-11:00	Discussion: Bayes-duality		Discussion: Influence estimation		Rebekka Burkholtz		CREST: Emti, Thomas, Nico		Razvan Pascanu
11:00-11:30	Coffee Break		Coffee Break		Coffee Break		Coffee Break		Coffee Break
11:30-12:30	Discussion: Bayes-duality		Discussion: Federated Learning		André Martins		CREST: Emti		Mark van der Wilk
12:30-14:00	Lunch		Lunch		Lunch		Lunch		Lunch
14:00-15:00	Discussion: Continual Leanring		Discussion: Model merging		David Rügamer		CREST: Thomas, Siddharth		Marcus Rohrbach
15:00-16:00	Discussion: LLM training		Free Discussion		Poster Session		CREST: Kenichi, Keigo, Cong Bai		Abeba Birhane
16:00-16:30	Coffee Break		Coffee Break		Coffee Break		Coffee Break		Coffee Break
16:30-18:00	Free Discussion		Free Discussion		Poster Session		Poster Session		Panel Discussion
18:00-	Dinner/Drinks						Speakers Dinner		Dinner/Drinks

Organizing Team

• Organizers: Emtiyaz Khan, Kenichi Bannai, Rio Yokota
• Webpage Design: Rio Yokota