Over the past decade, video games have catalyzed breakthroughs in reinforcement learning, yet as these important benchmarks become increasingly saturated, it is time to reflect on the lessons learned and explore what is still missing. What are the open research problems in RL that can be isolated and addressed through novel video game benchmarks? What is preventing RL from being more widely applied in modern video games?

To address these questions, the workshop focuses on two main themes:

Video Games for Benchmarking RL: How can modern video games serve as meaningful, targeted benchmarks to explore open problems in RL research, such as (but not limited to):

• How can we design RL agents to operate in vast, partially-observable open-world environments and engage in long-horizon tasks with sparse rewards?

• How can RL agents perform continual online learning when faced with novel tasks?

• How can RL agents collaborate or compete with human players, adapting to different playstyles and levels of expertise?

Leveraging RL for Video Games: How can insights from RL research drive innovation in video game design and development, such as (but not limited to):

• How can RL agents be trained as realistic opponents, NPCs, or QA agents in environments without clearly defined rewards?

• How can RL techniques generate novel and personalized content for production games?

• How can RL be leveraged in video games to improve player experience?

This workshop celebrates the 10th anniversary of the Deep Q-Network Atari Nature paper, a pivotal milestone that established video games as a cornerstone for RL research. By bringing together researchers using video games as platforms for RL and those incorporating RL into video game design, our goal is to critically evaluate the progress of the field over the past decade, identify gaps in current approaches, and chart a path forward with benchmarks that address both theoretical and practical challenges in RL.

Call for Paper: 15 April 2025

Workshop Paper Submission: 30 May 2025  6th June 2025

Notification of Results: 20 June 2025

Camera-ready Submission: 29 June 2025

Workshop Day: 5 August 2025

All times are 23:59 PM, AoE.

Email: rlvg2025 at gmail dot com 

Bluesky: @rlvgworkshop.bsky.social