Minqi Jiang
@MinqiJiang
Recently, there has been a lot of talk of LLM agents automating ML research itself. If Llama 5 can create Llama 6, then surely the singularity is just around the corner. How can we get a pulse check on whether current LLMs are capable of driving this kind of total…
Excited to release AlgoTune!! It's a benchmark and coding agent for optimizing the runtime of numerical code 🚀 algotune.io 📚 algotune.io/paper.pdf 🤖 github.com/oripress/AlgoT… with @OfirPress @ori_press @PatrickKidger @b_stellato @ArmanZharmagam1 & many others 🧵
Love this project: nanoGPT -> recursive self-improvement benchmark. Good old nanoGPT keeps on giving and surprising :) - First I wrote it as a small little repo to teach people the basics of training GPTs. - Then it became a target and baseline for my port to direct C/CUDA…
Recently, there has been a lot of talk of LLM agents automating ML research itself. If Llama 5 can create Llama 6, then surely the singularity is just around the corner. How can we get a pulse check on whether current LLMs are capable of driving this kind of total…
This is the most serious work I've seen on the path to recursive self-improvement (RSI), from Meta AI. It tasks agents with reproducing the chain of human innovations on improving LLMs. Proud that the authors used a scaffold extending AIDE (@WecoAI's core tech)!
Recently, there has been a lot of talk of LLM agents automating ML research itself. If Llama 5 can create Llama 6, then surely the singularity is just around the corner. How can we get a pulse check on whether current LLMs are capable of driving this kind of total…
The AIRA team @metaai has the ambitious goal of building/training an agent that can do frontier AI research to help the open-source ecosystem leapfrog closed source LLMs. As a relatively small team we cannot succeed in this mission without the support of the community so we'll…
Recently, there has been a lot of talk of LLM agents automating ML research itself. If Llama 5 can create Llama 6, then surely the singularity is just around the corner. How can we get a pulse check on whether current LLMs are capable of driving this kind of total…
A mental model I find useful: all data acquisition (web scrapes, synthetic data, RL rollouts, etc.) is really an exploration problem 🔍. This perspective has some interesting implications for where AI is heading. Wrote down some thoughts: yidingjiang.github.io/blog/post/expl…
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 @TimonWilli & @j_foerst at @AIatMeta & @FLAIR_Ox 🧵👇
You may not like it, but this is what a world-class AI + hardware team looks like.
The next few years of AI in a nutshell: Reinforcement learning with humans in the loop.
LLMs are both astonishing and brittle. The difference between superintelligence and stochastic parrot can be as small as a single whitespace character in your prompt.
Can you predict when the brain is processing speech/non-speech? The 2025 PNPL Competition 🏆 features two foundational decoding tasks which make efficient use of the PNPL🍍LibriBrain data. The first task to launch will be Speech Detection 🚀. More details: libribrain.com