Corentin LEGER
@corentinlger
Followers
137
Following
532
Media
4
Statuses
25
Research Engineer at @instadeepai Interested in RL, Multi-agent systems, ML & LLMs
Joined October 2022
🚀 Want to simulate #AI/#ALife agents in a physics-driven world? Meet Vivarium🌱– a new multi-agent simulator from @FlowersINRIA , designed for research & education! ✅ Large-scale simulations with #JAX ✅ User-friendly web interface ✅ Real-time control with jupyter notebooks
7
19
75
📢 New paper alert !! How to use Policy Gradient methods without explicit rewards? We address this question in our new work "From Data to Rewards: a Bilevel Optimization Perspective on Maximum Likelihood Estimation" 📜 https://t.co/mt0JhC7ZT7 🖥️ https://t.co/CtiqMyUt9q 1/🧵
1
3
8
3 papers from @FlowersINRIA are presented this week at the @ALifeConf , leveraging curiosity-driven AI to explore complex behaviors in #Lenia All first authors are attending the conference if you want to discuss with them Link to interactive websites, papers and code below:
2
8
15
Almost all agentic pipelines prompt LLMs to explicitly plan before every action (ReAct), but turns out this isn't optimal for Multi-Step RL 🤔 Why? In our new work we highlight a crucial issue with ReAct and show that we should make and follow plans instead🧵
5
41
172
🤔Why is tool use so effective for LLMs? In our new work, we provide theoretical and empirical evidence that tool-augmented workflows are not just practical but also provably more scalable. 📜 https://t.co/iBP1EDnt8T 🖥️ https://t.co/rxOnExsop7 1/🧵
1
6
10
🚀 NeurIPS@Paris is back for a 5th edition at the SCAI. Meet us in central Paris to discuss recent advances in AI! 📆 25th & 26th Nov. 2025 🌐 https://t.co/tgZJ1h2hIF 🎓 Committee: Chloé-Agathe Azencott, @BachFrancis, Claire Boyer, @gerardbiau , @VianneyPerchet , @jeanphi_vert
0
10
34
Introducing SOAR 🚀, a self-improving framework for prog synth that alternates between search and learning (accepted to #ICML!) It brings LLMs from just a few percent on ARC-AGI-1 up to 52% We’re releasing the finetuned LLMs, a dataset of 5M generated programs and the code. 🧵
8
38
192
🔔 Join our MAGELLAN talk on July 2! We'll explore how LLM agents can monitor their own learning progress and choose what to learn next, like curious humans 🤔 1h presentation + 1h Q&A on autotelic agents & more! 📅 July 2, 4:30 PM CEST 🎟️
docs.google.com
Speakers: Loris Gaven, Thomas Carta, Clément Romac Date: July 2, 2025 Time: 4:30pm (UTC+2) Duration: 2 hours (1h talk + 1h Q&A) In this talk, we will present MAGELLAN, a method that addresses a key...
🚀 Introducing 🧭MAGELLAN—our new metacognitive framework for LLM agents! It predicts its own learning progress (LP) in vast natural language goal spaces, enabling efficient exploration of complex domains.🌍✨Learn more: 🔗 https://t.co/uGLBSsOgMn
#OpenEndedLearning #LLM #RL
1
3
16
I'll be at RLDM this week to present our new paper: WorldLLM 😃 In the same spirit as our previous works — e.g. GLAM, MAGELLAN... — which investigate how to ground LLMs through interactions with external environments, WorldLLM takes a deep dive into the world modeling aspect.
2
6
18
I'm happy to share that MAGELLAN has been accepted at #ICML2025! 🎉 See you in Vancouver!
🚀 Introducing 🧭MAGELLAN—our new metacognitive framework for LLM agents! It predicts its own learning progress (LP) in vast natural language goal spaces, enabling efficient exploration of complex domains.🌍✨Learn more: 🔗 https://t.co/uGLBSsOgMn
#OpenEndedLearning #LLM #RL
0
4
15
Generative AI is a cultural transmission technology: it plays a growing role in generation, selection and transmission of ideas/opinions in human society 🧠🔄🌐 And yet we understand very little of this dynamics at this point 🤔❓ A step forward is our #ICLR2025 paper !👇
What happens when LLMs play the Telephone game? ☎️ In this new preprint, we analyse the evolution of texts as they are transmitted between LLM agents 🤖💬🤖💬🤖💬 Do text properties converge to attractors? 🧲 How is this influenced by the task📝 and model⚙️? 1/13🧵
1
6
39
Check out this work today at 10 am @iclr_conf! Learn what's wrong with evaluating #LLMs after a single interaction, and how cultural attraction theory can help us do better. Poster #288 @Jeremy__Perez @pyoudeyer @MaximeDerex @KovacGrgur @cedcolas @corentinlger @ClemMoulinFrier
What happens when LLMs play the Telephone game? ☎️ In this new preprint, we analyse the evolution of texts as they are transmitted between LLM agents 🤖💬🤖💬🤖💬 Do text properties converge to attractors? 🧲 How is this influenced by the task📝 and model⚙️? 1/13🧵
0
5
15
🔥Our paper PhyloLM got accepted at ICLR 2025 !🔥 In this work we show how easy it can be to infer relationship between LLMs by constructing trees and to predict their performances and behavior at a very low cost with @StePalminteri and @pyoudeyer ! Here is a brief recap ⬇️
3
13
33
🚀 Introducing 🧭MAGELLAN—our new metacognitive framework for LLM agents! It predicts its own learning progress (LP) in vast natural language goal spaces, enabling efficient exploration of complex domains.🌍✨Learn more: 🔗 https://t.co/uGLBSsOgMn
#OpenEndedLearning #LLM #RL
arxiv.org
Open-ended learning agents must efficiently prioritize goals in vast possibility spaces, focusing on those that maximize learning progress (LP). When such autotelic exploration is achieved by LLM...
3
11
38
👏 Built with @Clement_MF_ and Martial Marzloff! Explore Vivarium🌱here :
github.com
Multi-agent simulator in Jax for research and teaching in AI & ALife - flowersteam/vivarium
0
0
4
🔮 What’s next for Vivarium? - A more intuitive API is in development - Soon, you’ll be able to modularly compose diverse dynamics – imagine Braitenberg Vehicles interacting with Particle Lenia in just a few lines of code! 🎥 Check out the demo:
1
0
4
🎓 Teaching multi-agent systems? Vivarium makes it easy! ✅ Ready-to-use Jupyter notebook sessions for hands-on learning ✅ Covers key concepts like reactive behaviors & environmental dynamics ✅ Used in CISC Master's @UPFBarcelona 📚 Explore them here: https://t.co/v5cobZZyoE
1
0
4
🎮 What can you do with Vivarium? - Simulate agents with sensors & motors inspired by Braitenberg Vehicles - Make them interact in a 2D physics world (built with JAX & JAX-MD) - Visualize & control simulations dynamically in real time 🔗 Try it here: https://t.co/BjMgmO0fyN
1
0
7
Excited to share with @flowersINRIA ADTool, an open-source Python toolkit using curiosity-driven AI to explore behavior diversity of complex systems (cellular automata, physics models, ...)!🐍 Serving off-the-shelf strategies like IMGEP https://t.co/UGl6VhhH3A You said IMGEP? 👇
2
27
112
🚨 NEW LEADERBOARD ! 🚨 We instruct LLMs to role-play and then add a long context. Which LLMs best Stick to Their Role? Imagine asking several LLMs to play a role, then discuss various topics. Which ones best stick to their role? 🥇 Mistral-Large 🥈 Qwen2.5-72B 🥉 Llama3.1_70b 🧵
7
18
48