Finding the Frame

An RLC Workshop for Examining Conceptual Frameworks in RL

August 09, 2024

About

Scientific research is often framed within paradigms of concepts, definitions, axioms, and assumptions — RL is no exception. While such details usually go unchallenged in the pursuit of specific objectives, this workshop brings them into sharp focus. In this workshop, we examine the conceptual foundations of RL and their impact on framing problems. We are excited for this workshop, because we believe it has the potential to open up new territory, test our historical beliefs, and ultimately guide our future research endeavors in productive directions. Overall, we hope to generate enthusiasm and build a community dedicated to examining the conceptual foundations of RL.

This is a full-day workshop. The first half of the day features four short talks, representing different perspectives on RL. Following these talks, these speakers join a panel discussion. Using this panel as backdrop for the rest of the day, we open the floor to submitted work, which includes poster sessions and lightning talks. (See the detailed schedule.)

Call for Papers

Topics. We invite submission of papers of up to six content pages (excluding the references, acknowledgements, and appendices) that reflect on the philosophy, practice, and formalisms of reinforcement learning. This includes, but is not limited to, the following areas:

Proposals that explore beyond the standard RL problem or paradigm.
Any work rethinking goals, models, values, or rewards in the context of RL.
Methods, benchmarks, or applications that explore limitations of our previous paradigms.
Reflections on the basic building blocks or methods of reinforcement learning.
Issues with existing challenges and benchmarks that artificially or deceptively narrow the RL frame.
Work that investigates formulations of the RL problems that are implicit in our solution methods.

We also welcome submissions that interpret our theme in other creative ways.

Submission Instructions. Submissions will be made through OpenReview, formatted using the RLC author kit (overleaf link or zip file link). All submissions will be reviewed double-blind, with accepted abstracts posted on the workshop website. Accepted submissions will be presented at two poster sessions during the workshop. Time depending, selected works will be highlighted during lightning talks.