Chris Lu (@_chris_lu

Chris Lu

340 posts

Chris Lu

@_chris_lu_

Research @OpenAI Prev: DPhil Student @UniofOxford, RS Intern @SakanaAILabs @DeepMind and RS @CovariantAI

Joined November 2014

Pinned
Chris Lu
@_chris_lu_
Aug 13, 2024
Excited to share The AI Scientist! We use LLMs to autonomously come up with research ideas, implement them, do literature search, write them up, and review them -- producing full-length papers on AI without human intervention. Co-led with @cong_ml and @RobertTLange
Sakana AI
@SakanaAILabs
Aug 13, 2024
Introducing The AI Scientist: The world’s first AI system for automating scientific research and open-ended discovery! sakana.ai/ai-scientist/ From ideation, writing code, running experiments and summarizing results, to writing entire papers and conducting peer-review, The AI
73K
Chris Lu
@_chris_lu_
Aug 5, 2025
To all my academic friends who gave me crap for joining OpenAI: We just open-sourced some banger models. Have fun with them!
OpenAI
@OpenAI
Aug 5, 2025
Our open models are here. Both of them. openai.com/open-models
178K
Chris Lu
@_chris_lu_
Apr 6, 2023
1/ 🚀 Presenting PureJaxRL: A game-changing approach to Deep Reinforcement Learning! We achieve over 4000x training speedups in RL by vectorizing agent training on GPUs with concise, accessible code. Blog post: chrislu.page/blog/meta-disc… 🧵
165K
Chris Lu
@_chris_lu_
Nov 20, 2023
Crazy times. Anyways, excited to unveil JaxMARL! JaxMARL provides popular Multi-Agent RL environments and algorithms in pure JAX, enabling an end-to-end training speed up of up to 12,500x! Co-led w/ @alexrutherford0 @benjamin_ellis3 @MatteoGallici Post: blog.foersterlab.com/jaxmarl/
67K
Chris Lu
@_chris_lu_
Mar 16, 2023
Blazingly-fast sequence models enable Meta-RL agents that generalise across a wide range of different tasks! I'm excited to share my @DeepMind internship project, where we look at applying recent advancements in SSM’s to in-context RL! Link: arxiv.org/abs/2303.03982 🧵
GIF
34K
Chris Lu
@_chris_lu_
Nov 23, 2022
Deep RL has been driven by improvements in handcrafted algorithms. Our NeurIPS 2022 paper, “Discovered Policy Optimisation” instead meta-learns in a space of theoretically-sound algorithms and beats PPO on unseen tasks! w/ @kuba_AI @_aletcher @Luke_Metz @casdewitt @j_foerst 🧵
Chris Lu
@_chris_lu_
Jun 13, 2024
Excited to share my first work from my internship @SakanaAILabs! We used LLMs to design and implement new preference optimization algorithms for training LLMs, discovering cutting-edge methods! Co-led with @samianholt and Claudio Fanconi. Details in thread 🧵 (1/N)
Sakana AI
@SakanaAILabs
Jun 13, 2024
Can LLMs invent better ways to train LLMs? At Sakana AI, we’re pioneering AI-driven methods to automate AI research and discovery. We’re excited to release DiscoPOP: a new SOTA preference optimization algorithm that was discovered and written by an LLM! sakana.ai/llm-squared/
00:00
38K
Chris Lu
@_chris_lu_
Feb 20, 2024
How can we meta-learn new RL algorithms that vastly outperform PPO and its variants? In our ICLR 2024 paper, we find that *temporally-aware* algorithms unlock performance gains, significantly beating PPO on unseen tasks! work co-led with @JacksonMattT at @whi_rl @FLAIR_Ox
GIF
16K
Chris Lu
@_chris_lu_
Jun 19, 2023
🚀 Excited to announce our ICML 2023 paper "Adversarial Cheap Talk", which has deep implications for AI safety in RL (such as RLHF or recommender systems)! We show that an Adversary can manipulate an RL agent's performance and test-time behavior with *minimal* access.
GIF
20K
Chris Lu
@_chris_lu_
Dec 12, 2023
If you're at NeurIPS, swing by Poster #1226 at 17:15 to discuss our paper on applying SSM's (like Mamba) to RL! A bonus: We've released a pure JAX repo for testing these architectures that includes 34 memory-based environments and runs 100x faster github.com/luchris429/pop…
11K
Chris Lu
@_chris_lu_
Apr 3, 2024
If you're interested in any type of RL research, you have to try end-to-end pure JAX. With all of these new libraries and features coming out, theres no better time to dive in. Great work! @scowardai @mcbeukman
Sam Coward
@scowardai
Apr 2, 2024
🔧 Looking to easily develop RL novel autocurricula methods? ⚡ Want clean, blazingly fast, and readily-modifiable baselines? Presenting JaxUED: a simple and clean RL autocurricula library written in Jax!
GIF
5.3K
Chris Lu
@_chris_lu_
Jan 12, 2024
Pure Jax achieves up to 250x speed up for Inverse RL! Run your experiment in 3.5 minutes instead of ~15 hours. Check out our new repository!
Silvia Sapora
@silviasapora
Jan 12, 2024
1/🧵Jax fans, buckle up! Prepare to ditch slowpoke IRL training with the ⚡️jaxirl⚡️ launch, a JAX library built for Inverse Reinforcement Learning algorithms that's 250x faster 🏎️ 💨 than PyTorch. Read on to learn more ⬇️⬇️⬇️
GitHub - FLAIROx/jaxirl: Contains JAX implementation of algorithms for inverse reinforcement...
From github.com
8.1K
Chris Lu
@_chris_lu_
Nov 20, 2023
Replying to @_chris_lu_
We implement several MARL environments in pure JAX, including Hanabi, Overcooked, Coin Game, and MPE! We also introduce MABrax (based on MAMujoco), SMAX (based on SMACv2), and STORM (based on Melting Pot). Our envs run many thousand times faster than their CPU-based counterparts
24K
Chris Lu
@_chris_lu_
Jun 6, 2024
AI has progressed by learning from human data, so how can it surpass human intellect? Our pre-print explores how cultural accumulation—the secret to the our species’s success—could emerge in artificial learning agents. Co-led with @JonnyCoook. arXiv:
Jonny Cook
@JonnyCoook
Jun 6, 2024
1/ 🚀 Presenting AGI - Artificial Generational Intelligence 🚀 We apply the concept of cultural accumulation to RL and find that agents can improve across generations, outperforming those trained for a single lifetime of the same experience budget! Co-led w/ @_chris_lu_. 🧵
arxiv.org
Artificial Generational Intelligence: Cultural Accumulation in...
Cultural accumulation drives the open-ended and diverse progress in capabilities spanning human history. It builds an expanding body of knowledge and skills by combining individual exploration...
7.3K