🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Excited to announce Jumanji v1.0, now featuring 22 fast, flexible, and scalable environments! Fully written in JAX, Jumanji offers on-device fully-jitted simulations and training. Jumanji got published at ICLR 2024! Paper: arxiv.org/abs/2306.09884 GitHub: github.com/instadeepai/ju…

If you haven't yet, please check out the amazing works from the JAX community. E.g. environments: Brax (Olivier Bachem), Gymnax (Robert Lange), JaxMARL (Alex Rutherford), Craftax (Michael Matthews), Pgx (Sotetsu K.（そーてつ）). And algorithms: PureJaxRL (Chris Lu), Stoix/Flashbax (Edan Toledo).

Curious about this diagram? Join us later today as we discuss growing the MARL ecosystem in JAX! 🤖🍿 InstaDeep Ruan de Kock Omayma Mahjoub Sasha @formanek_claude (& for a sneak preview: arxiv.org/abs/2107.01460 😉)

Got lost in the #ICLR2024 poster maze? Don't worry, we've got your covered! 🛟 Here is Donal Byrne, Senior Research Engineer at InstaDeep , as he showcases Jumanji — our library for high-performance RL environments in #JAX ⭐️ Github: tinyurl.com/code-jumanji

Excited to introduce our latest neural solver, MEMENTO! Enhancing problem-specific adaptation with an explicit memory. Thanks to my InstaDeep collaborators: Refiloe 🇱🇸, Noah🇿🇦, Arnu Pretorius🇷🇼, Tom Barrett 🇬🇧, Nathan Grinsztajn🇬🇧! arxiv.org/abs/2406.16424 🧵[1/9]

What happens when trying to learn multi-agent coordination from a static dataset? Catastrophe, if you’re not careful! This is the topic of our latest work on ✨Coordination Failure in Offline Multi-Agent Reinforcement Learning ✨ Curious about this image? Read below 👇 [1/16]

Can we fix the review process before we try to automate science?

It's amazing two of the 2024 #NobelPrize were for AI! But as they say: it took a village. "We didn't win a Nobel", a parody of Billy Joel 's "We didn't start the fire" covers a tiny sliver of this historical "village". Hope you enjoy it as much as I did making it!

Introducing Latent Program Network (LPN), a new architecture for inductive program synthesis that builds in test-time adaption by learning a latent space that can be used for search 🔎 Inspired by ARC Prize 🧩, we designed LPN to tackle out-of-distribution reasoning tasks!

Excited to share our latest work on Sequential Monte Carlo Policy Optimisation (SPO)🔥— a scalable, search-based RL algorithm leveraging SMC as a policy improvement operator for both continuous and discrete environments! 📍 Catch us tomorrow at #NeurIPS2024 (poster #94776) from

Efficient LLM reasoning over large data doesn't require massive contexts! 🫡 We show that a simple in-context method, PRISM, allows a 32k token model to outperform baselines and sometimes rival a 1M token model while saving up to 54% on token cost. w/ Google DeepMind

Thrilled to see our NeurIPS 2024 paper, Sequential Monte Carlo Policy Optimisation (arxiv.org/abs/2402.07963), featured in Kevin's Reinforcement Learning: A Comprehensive Overview, which additionally recognises SMC as a competitive, scalable online planner. A fantastic modern

New paper accepted to ICML! We present a novel policy optimization algorithm for continuous control with a simple closed form which generalizes DDPG, SAC etc. to generic stochastic policies: Wasserstein Policy Optimization (WPO).

Theory of Mind (ToM) is crucial for next gen LLM Agents, yet current benchmarks suffer from multiple shortcomings. Enter 💽 Decrypto, an interactive benchmark for multi-agent reasoning and ToM in LLMs! Work done with Timon Willi & Jakob Foerster at AI at Meta & Foerster Lab for AI Research 🧵👇

AI Research Agents are becoming proficient at machine learning tasks, but how can we help them search the space of candidate solutions and codebases? Read our new paper looking at MLE-Bench: arxiv.org/pdf/2507.02554 #LLM #Agents #MLEBench

Very proud of this work! If you're interested in AI agents and their current challenges, give this a read. Thanks to my incredible collaborators and to Meta and UCL for enabling me to tackle something of this scale for my first PhD paper. Excited for what's ahead!