Ludovic Denoyer

Hi. We propose two research internships this year focused on bot development @UbisoftLaForge : * Learning distributions of policies to capture the diversity of players behaviors jobs.smartrecruiters.com/Ubisoft2/74399… * Continual and Transfer RL for bots: jobs.smartrecruiters.com/Ubisoft2/74399… 1/3

First days in the new @imec_int Paris Office where we are building a new frontier AI research team together with @karl_tuyls . A little bit empty right now 😄 but researchers and engineers are joining in the next days and weeks. Super exciting time 🙃

🏄 Our Surfer-H agent, powered by our Holo1 model, hits 92.2% SOTA on WebVoyager! Achieves Pareto-optimal accuracy & cost-efficiency, outperforming GPT-4.1 and other foundation models at a fraction of the cost. hcompany.ai/surfer-h

Le Sommet de Paris sur l'IA sera co-présidé avec l'Inde 🇮🇳 Je regrette cette décision tout comme une bonne partie du programme annoncé. Ne cherchez pas @WiMLDS_Paris. Visiblement, ils ont perdu notre adresse / oublié notre existence. Ce sera sans nous maddyness.com/2025/01/16/som…

Very excited to share what the team has been cooking up at @hcompany_ai since May. We are 100% focused on building the agentic stack, including specialized action models as well as the agents that run these models, all supporting our product. Learn more about our research here hcompany.ai/blog/a-researc…

@RemiCadene La classe !

Le monde de la tech française est à l'écoute des avancées vertigineuses en robotique 🦾 Le moment ChatGPT de la robotique est pour bientôt 🚀

@clongmire42 Nice work. But one of the questions is about the time spent to setup the two approaches. If one is cheaper, at a similar level of quality, then it is better.

If you can't spot a big difference with run, here's walk. If your answer is "wow, they look very similar", than you're right! The lesson here is that motion matching and state machines are merely just tools. In the right hands, great results can be achieved with both.

Here's a thing I've been passionately working away on: What if I told you that one of these characters is using a Motion Matching setup, and the other is using a simple State Machine (sort of)? Would you believe me? Take a guess at which is which.

@francoisfleuret @Meta @syhw @unige_en @EPFL Ouah ! Congrats

Some career update! I am joining @meta (FAIR) to work in @syhw's team, I remain Prof part-time at @unige_en, and my affiliation with @epfl comes to an end. 🎶 We are going to make AI 🎶

You are a M2 student looking for a reseach internship in neural rendering? Come work with us at @UbisoftLaForge (based in Bordeaux). Neural Textures for Complex Materials: jobs.smartrecruiters.com/Ubisoft2/74400… Neural Representation of Large-Scale Environments: jobs.smartrecruiters.com/Ubisoft2/74400…

@kohler_hector @scikit_learn Related ? arxiv.org/abs/1505.00908

@francoisfleuret It makes me remember of this old paper we wrote in 2009 about sequence labelling in any order. At that time, tasks were simpler :) academia.edu/download/43948…

Here comes σ-GPT! TL;DR: We added an output positional encoding, which allows to generate tokens in any order chosen on the fly. Tweet 5/6 is sweet.

Team Yann 😀

(Longer version: linkedin.com/posts/ludovic-…)

I think that gaming has to accelerate the adoption of open-sourcing, massive (anonymized) players data tracking, benchmarking (e.g A/B testing) and fast iterations, which made AI successful in other industries. Pushing in that direction was sometimes a friendly clash of cultures

Two years ago, I decided to leave Meta and join Ubisoft. It was a return to the core passion that initially drew me to AI long before the field's expansion. I had the privilege of joining the new extension of @UbisoftLaForge in Bordeaux, led by (the fantastic) Guillaume Gaudron

Today, I visited the robotic lab started by my former student and friend @RemiCadene at @huggingface in Paris. Very nice to see such an initiative to promote open-source robotics on both hardware and software sides !

When the only poor people you meet are the GPU-poor

Congratulations to @jb_gaya who defended last week his thesis entitled "Subspaces of Policies for Deep Reinforcement Learning" (Co-supervised with @LudovicDenoyer and Alessandro Lazaric) It was a pleasure to supervise your thesis, I wish you all the best for your future!