Two RL papers accepted at ICLR 2026 : * Learning a subspace of policies for online adaptation in RL - with Jean-Baptiste Gaya and Laure Soulier - tinyurl.com/5eh6z5uk * Direct then Diffuse: Incremental Unsupervised Skill Discovery ... - with @pa_kamienny Jean Tarbouriech Alessandro Lazaric

Our work on learning exploration strategies without extrinsic rewards was accepted at ICLR 2026! We learn a tree-structured policy that composes skills to reach increasingly far states. Work with Jean Tarbouriech Alessandro Lazaric Ludovic Denoyer. Camera-ready paper coming soon.

Towards a better understanding of unsupervised goal-conditioned RL in our #AISTATS2022 paper! Join us at poster session 4 on Wed, Mar 30 (10:30-12 UTC) AISTATS Conference Adaptive Multi-Goal Exploration arxiv.org/pdf/2111.12045… w/ Omar D. Pierre Ménard Matteo Pirotta Michal Valko Alessandro Lazaric

I’m super excited to share our work on AdA: An Adaptive Agent capable of hypothesis-driven exploration which solves challenging unseen tasks with just a handful of experience, at a similar timescale to humans. sites.google.com/corp/view/adap… See the thread for more details 👇 [1/N]

Excited to present the 'epistemic-risk-seeking actor-critic' - an actor-critic / policy gradient algorithm that performs deep exploration. arxiv.org/abs/2302.09339 Accepted at #ICML2023

I'm super excited to share AlphaZeroᵈᵇ, a team of diverse #AlphaZero agents that collaborate to solve #Chess puzzles and demonstrate increased creativity. Check out our paper to learn more! arxiv.org/abs/2308.09175 A quick 🧵(1/n)

Really excited about this work! We consider the notion of Bayesian optimality in RL. Computing the exact posterior is computationally intractable, so we derive a new variational approximation that maintains the same low regret. Great work led by Jean Tarbouriech !

Excited to share what my team has been working on lately - Gemini diffusion! We bring diffusion to language modeling, yielding more power and blazing speeds! 🚀🚀🚀 Gemini diffusion is especially strong at coding. In this example the model generates at 2000 tokens/sec,

The Future of Development: Gemini Diffusion

We're looking for people to join us to work on Gemini Diffusion and help revolutionize language modeling! Details below: job-boards.greenhouse.io/deepmind/jobs/…