Maxime Peyrard
peyrardmax.bsky.social
Junior Professor CNRS (previously EPFL, TU Darmstadt) -- AI Interpretability, causal machine learning, and NLP. Currently visiting @NYU https://peyrardm.github.io
👤 Joined on 18 November 2024
602
Followers
178
Following
12
Total Posts
📜 Kullanıcı Geçmişi
- 18 Kas 2024, 15:30
🎉 Bluesky'e katıldı - İlk kullanıcı adı:
@peyrardmax.bsky.social
🚫 Blokladığı Kullanıcılar
❤️ Beğendiği Gönderiler
Work done with great teams at @naverlabseurope.bsky.social @getalp.bsky.social @peyrardmax.bsky.social See you there! #EMNLP2026 #LLMAgents #RAG #AgentMemory #MechanisticInterpretability #LLMReasoning #LatentReasoning
Show Post🧠 Retrieval-Augmented LLM Agents: Learning to Learn from Experience Simple episodic experience retrieval improves generalization to unseen tasks, especially when added to training, and matches / beats agent memory systems. [Major updates from review discussion coming soon] arxiv.org/abs/2603.18272
Show Post🔍 Observable Patterns Are Not Explanations: A Causal-Geometric Analysis of Latent Reasoning Models Latent thoughts are hidden computation, not explanations. Causality should come first, then we can interpret the geometry of the revealed behavior-influential spaces. arxiv.org/abs/2606.12689
Show Post🎉 Two papers accepted to #EMNLP2026 Main! 1️⃣ Studying agent retrieval and learning from experience, we find it can be as good as overcomplex memory. 2️⃣ Revisit and contest mechanistic claims from latent reasoning models and discuss its interpretability. See you in Budapest! 🇭🇺 Links / TLDRs 👇🧵
Show Post
I'm very happy to give a spotlight at the Mechanistic Interpretability Workshop @ ICML on our work: "Validating Causal Abstraction Metrics on Simulated Complex Systems" Which metrics actually tell you if an explanation is valid? We built a benchmark to find out. 1/n
Show Post