Florent Delgrange (@f_delgrange) 's Twitter Profile
Florent Delgrange

@f_delgrange

postdoc @aibrussels; working on providing reliable and verifiable AI mechanisms, with a strong focus on Reinforcement Learning

ID: 1647952411279892480

linkhttp://delgrange.me calendar_today17-04-2023 13:17:44

26 Tweet

71 Followers

271 Following

Gaspard Lambrechts (@gsprdlambrechts) 's Twitter Profile Photo

Today we had the pleasure of receiving Florent Delgrange and⁩ ⁦Raphael Avalos⁩ for exciting talks on Wasserstein Auto-Encoded MDPs and it’s application to POMDP with the Wasserstein Believer! WAE-MDP: openreview.net/pdf?id=JLLTtEd… Wasserstein Believer: openreview.net/pdf?id=V5GQVp8…

Today we had the pleasure of receiving <a href="/f_delgrange/">Florent Delgrange</a> and⁩ ⁦<a href="/raphael_avalos/">Raphael Avalos</a>⁩ for exciting talks on Wasserstein Auto-Encoded MDPs and it’s application to POMDP with the Wasserstein Believer! 

WAE-MDP: openreview.net/pdf?id=JLLTtEd…

Wasserstein Believer: openreview.net/pdf?id=V5GQVp8…
Raphael Avalos (@raphael_avalos) 's Twitter Profile Photo

Damien ERNST Thanks a lot for receiving Florent Delgrange and me to present our research on POMDPs. We had a great time chatting with you and your team Gaspard Lambrechts , Pascal Leroy and the others! šŸ“„arxiv.org/abs/2303.03284

Pablo Samuel Castro (@pcastr) 's Twitter Profile Photo

Yes! Bisimulation FTW! Some papers to get you started: Bisim relations: sciencedirect.com/science/articl… Bisim metrics: arxiv.org/abs/1207.4114 Using these in deep RL: arxiv.org/abs/2106.08229

Raphael Avalos (@raphael_avalos) 's Twitter Profile Photo

Arrived at #ICLR2024 with Florent Delgrange to present our work "The Wasserstein Believer: Learning Belief Updates for Partially Observable MDPs through Reliable Latent Space Models".

Arrived at #ICLR2024 with <a href="/f_delgrange/">Florent Delgrange</a> to present our work "The Wasserstein Believer: Learning Belief Updates for Partially Observable MDPs through Reliable Latent Space Models".
Willem Rƶpke (@willem_ropke) 's Twitter Profile Photo

Exciting news! My paper on multi-objective reinforcement learning was accepted at AAMAS 2025! We introduce IPRO (Iterated Pareto Referent Optimisation)—a principled approach to solving multi-objective problems. šŸ”— Paper: arxiv.org/abs/2402.07182 šŸ’» Code: github.com/wilrop/ipro

Guy Avni (@guyavni) 's Twitter Profile Photo

Sit back, relax, and let me tell you the story of our paper in @aamasconf, by far the longest project I’ve been involved in; starting in the days of yore, just weeks before COVID hit. With F. Delange, C. Schilling, A. Lukina, A. Nowe, and G. Perez. arxiv.org/abs/2402.13785

Raphael Avalos (@raphael_avalos) 's Twitter Profile Photo

Last week, I wrapped up my internship cohere, where I had the chance to work with fantastic people on RL for LLMs. It was an amazing 6 months, and I'm excited to share one of the outcomes: ShiQ, a Q-value based RL algorithm for fine-tuning LLMs šŸš€ 🧵Details in Irem Ergün's post!