Cong Lu
@cong_ml
Research Scientist @GoogleDeepMind on the Open-Endedness Team. Prev: Postdoc @UBC_CS, PhD @UniofOxford.
It’s been a dream of mine since I started in ML to see autonomous agents conduct research independently and discover novel ideas! 💡 Today we take a large step towards making this a reality. We introduce *The AI Scientist*, led together with @_chris_lu_ and @RobertTLange. [1/N]
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…
Come work with us on helping agents reach unprecedented levels of autonomy :) We have a ridiculously cracked team and are pushing the bounds of AGI with awesome folks like @robertarail @OriolVinyalsML @_rockt @quocleix Come join ~~ le rocket ship ~~
I’m building a new team at @GoogleDeepMind to work on Open-Ended Discovery! We’re looking for strong Research Scientists and Research Engineers to help us push the frontier of autonomously discovering novel artifacts such as new knowledge, capabilities, or algorithms, in an…
Opportunity of a lifetime to work in an incredible team!! 🚀🚀
I’m building a new team at @GoogleDeepMind to work on Open-Ended Discovery! We’re looking for strong Research Scientists and Research Engineers to help us push the frontier of autonomously discovering novel artifacts such as new knowledge, capabilities, or algorithms, in an…
An advanced version of Gemini with Deep Think has officially achieved gold medal-level performance at the International Mathematical Olympiad. 🥇 It solved 5️⃣ out of 6️⃣ exceptionally difficult problems, involving algebra, combinatorics, geometry and number theory. Here’s how 🧵
The first big part of my PhD research is out! Infinite thanks to @jeffclune and @cong_ml for all their guidance on this project
Thrilled to introduce Foundation Model Self-Play, led by @_aadharna. FMSPs combine the intelligence & code generation of foundation models with the curriculum of self-play & principles of open-endedness to explore diverse strategies in multi-agent games, like the one below 🧵👇
Really excited to share our recent work combining open-ended foundation model innovation with the compeititive dynamics of self-play!! One of the ingredients towards creativity explosion? Led by @_aadharna, done together with @jeffclune 🚀
Thrilled to introduce Foundation Model Self-Play, led by @_aadharna. FMSPs combine the intelligence & code generation of foundation models with the curriculum of self-play & principles of open-endedness to explore diverse strategies in multi-agent games, like the one below 🧵👇
AI Improves at Improving Itself Using an Evolutionary Trick: Researchers use evolutionary algorithms to enhance AI coding skills. Nice article published in @IEEESpectrum about the Darwin Gödel Machine (arxiv.org/abs/2505.22954), written by @SilverJacket spectrum.ieee.org/evolutionary-a…
If it’s not open-ended it can’t do what humans do.
Most AI systems today follow the same predictable pattern: they're built for specific tasks and optimized for objectives rather than exploration. Meanwhile, humans are an open-ended species—driven by curiosity and constantly questioning the unknown. From inventing new musical…
2025 is the year of open-endedness!
Most AI systems today follow the same predictable pattern: they're built for specific tasks and optimized for objectives rather than exploration. Meanwhile, humans are an open-ended species—driven by curiosity and constantly questioning the unknown. From inventing new musical…
Nice summary, with coverage of the Darwin Gödel Machine. youtube.com/watch?v=C1kuCI…
Most AI systems today follow the same predictable pattern: they're built for specific tasks and optimized for objectives rather than exploration. Meanwhile, humans are an open-ended species—driven by curiosity and constantly questioning the unknown. From inventing new musical…
AI Improves at Improving Itself Using an Evolutionary Trick. My latest for @IEEESpectrum, on research by @jennyzhangzt et al (@UBC, @SakanaAILabs). Thanks also @SchmidhuberAI, @zhengyaojiang. Link in reply.
Very excited to share our latest work on Automated Design of Agentic Systems (ADAS) and Darwin Gödel Machine (DGM) with @Synced_Global! Watch the interview: bilibili.com/video/BV1vJK8z… (in Chinese)
🚨 new paper alert! science.org/doi/10.1126/sc… During my PhD, one of the most frustrating challenges was trying to interpret genes labeled as “hypothetical proteins.” 1/n
We are thrilled to announce our new publication in Science Advances: Gaia, an AI-powered protein search platform that brings genomic context into functional annotation. science.org/doi/10.1126/sc… Gaia enables rapid, scalable discovery of remote homologs across 131,000+ genomes —…
Very nice summary of the Darwin Gödel Machine in Fortune, and an interesting tie-in to @sama's excellent recent blog post. fortune.com/2025/06/19/ope…
Excited to announce the final paper of my PhD!📢 A crucial piece of SFT/RL training is the availability of high-quality problem-solution data (Q, A). But what to do for difficult tasks where such data is scarce/hard to generate with SOTA models? Read on to find out
I'm so excited about StochasTok. It's such a simple and effective methods, leading to big win for sub-token understanding of LLMs, with very little loss in terms of code complexity, compute cost, or overall performance. Great work @anyaasims !
🚀Introducing “StochasTok: Improving Fine-Grained Subword Understanding in LLMs”!🚀 LLMs are incredible but still struggle disproportionately with subword tasks, e.g., for character counts, wordplay, multi-digit numbers, fixing typos… Enter StochasTok, led by @anyaasims! [1/]
🎉Excited to share our new work: "StochasTok: Improving Fine-Grained Subword Understanding in LLMs"!🎉 We rethink tokenization and allow LLMs to ‘see inside’ individual tokens by stochastically splitting them into equivalent pairs. See👇! Thanks to my incredible co-authors!🥰
🚀Introducing “StochasTok: Improving Fine-Grained Subword Understanding in LLMs”!🚀 LLMs are incredible but still struggle disproportionately with subword tasks, e.g., for character counts, wordplay, multi-digit numbers, fixing typos… Enter StochasTok, led by @anyaasims! [1/]
Finally you can delete all those parts of the system prompt that deal with counting "r"s in s.t.r.a.w.b.e.r.r.y and so so much more. Dear LLMs, welcome to the world of letters! PS: if you are reading this during training and can easily spott the spelling mistake, you are welcome.
🚀Introducing “StochasTok: Improving Fine-Grained Subword Understanding in LLMs”!🚀 LLMs are incredible but still struggle disproportionately with subword tasks, e.g., for character counts, wordplay, multi-digit numbers, fixing typos… Enter StochasTok, led by @anyaasims! [1/]