Last week, members of our group, Carolin Schmidt, Miguel Costa, Arthur Vandervoort, and former member Daniele Gammelli had the honor of presenting at ICLR2025 in Singapore!

Carolin and Daniele presented their work titled “OHIO: Offline Hierarchical Reinforcement Learning via Inverse Optimization”— a new framework that allows you to use off-the-shelf offline RL algorithms to effectively learn hierarchical policies from arbitrary datasets. You can find more on the website 🌐: https://ohio-offline-hierarchical-rl.github.io/

Miguel and Arthur presented their latest work from MA’AT Project titled “Using Reinforcement Learning to Integrate Subjective Wellbeing into Climate Adaptation Decision Making,” which received the Best Proposal Award at the Tackling Climate Change with Machine Learning workshop. This work proposed a novel framework for integrating wellbeing into climate adaptation planning—specifically addressing urban flooding—using reinforcement learning. If you want to know more about this project or our short paper, you can find more information here: https://linktr.ee/_maat

💡 A huge thank you to all who visited their poster sessions and contributed to insightful discussions.