Umberto Lupo

DiffPaSS: Differentiable Pairing using Soft Scores Optimize pairings and graph alignments using gradients and smooth scores 📜 @gembioworkshop @iclr_conf paper: tinyurl.com/2p98cvrr 🖥️ #PyTorch library: bitbol-lab.github.io/DiffPaSS/ Specialised to interacting #protein sequences

Many of the coolest people here now post there regularly, in some cases *exclusively*. So you won't miss important science announcements, and increasingly might miss some cool content/discussions if you remain here. I will no longer post, comment or react on this platform.

I hope to see many of you soon in the other place 🦋. The science community there (and the BioML one in particular) looks much, much more lively now than last time I checked!

The US election is nigh. If you RAGE that it’s down to a few votes in a few swing states, I've got a story for you. A political drama about just how close we came to abolishing the Electoral College. And BONUS, it’ll explain why we still use this old relic to pick Presidents.🧵

@alexmd2 @skdh I am a random person on the internet, and am with @skdh on this one. I much prefer to refer to "Americans" as the inhabitants of the American continents, despite being in the minority here. en.wikipedia.org/wiki/Demonyms_…

@skdh What is US-American? It’s just Americans.

US-Americans like to brag about how great their democracy is, alas, according to the Economist Democracy Index (2023) the USA is not even ranked a full democracy any more, but a "flawed democracy", well behind most European countries (and Canada).

circles encircling circles... (this took an entire day to render... 🥲) (it's ya boi, the hopf fibration of the 3-sphere...)

#MLCB2024 kicks off in 22h (9am PST) with keynote from @zhou_jian on DL for transcriptional initiation! Full schedule at mlcb.github.io and public livestream at youtube.com/live/reqWvNOKl…. Please RT!

@sara_mostafavi Congrats on the decision to live stream this!

#MLCB2024 will be streamed live from YouTube this week (Thursday and Friday). Get the link from here: mlcb.github.io

🚀Excited to announce: Open-source AlphaFold3 implementation! 🚀 I am thrilled to announce one of the models we have been building for the last 8-weeks at Ligo - an open-source implementation of DeepMind’s frontier model, AlphaFold3! Here’s what we have learned, a thread (1/11):

"24 of the brain samples [...] measured on average about 0.5% plastic by weight." "[...] 12 brain samples from people who had died with dementia [...] contained up to 10 times more plastic by weight than healthy samples." Cool! 📜ncbi.nlm.nih.gov/pmc/articles/P… theguardian.com/environment/ar…

The etymologist X needs, but not the one it deserves right now

EPFL Faculty Position Life Science Engineeering: epfl.ch/about/working/…

Excited to see this. Looking forward to even tighter integration at some point.

Organoids are amazing

I support Kamala because she has amazing music tastes and her step-kids are named after Coltrane

There was a super impressive AI competition that happened last week that many people missed in the noise of AI world. I happen to know several participants so let me tell you a bit of this story as a Sunday morning coffee time. You probably know the Millennium Prize Problems where the Clay Institute pledged a US$1 million prize for the first correct solution to each of 7 deep math problems. To this date only one of these, the Poincaré conjecture, has been solved by Grigori Perelman who famously declined the award (go check Grigori out if you haven't the guy has a totally based life). So this new competition, the Artificial Intelligence Math Olympiad (AIMO) also came with a US$1M prize but was only open to AI model (so the human get the price for the work of the AI...). It tackle also very challenging but still simpler problems, namely problems at the International Math Olympiad gold level. Not yet the frontier of math knowledge but definitely above what most people, me included, can solve today. The organizing committee of the AIMO is kind-of-a who-is-who of highly respected mathematicians in the world, for instance Terence Tao widely famous math prodigy widely regarded as one of the greatest living mathematicians. Enter our team, Jia Li, Yann Fleuret, and Hélène Evain. After a successful exit in a previous startup (that I happen to have know well when I was an IP lawyer in a previous life but that's for another story) they decided to co-found Numina as a non-profit to do open AI4Math. Numina wanted to act as a counterpoint to AI math efforts like DeepMind's but in a much more open way with the goal to advance the use of AI in mathematics and make progress on hard, open problems. Along the way, they managed to recruit the help of some very impressive names in the AI+math world like Guillaume Lample, co-founder of Mistral or Stanislas Polu, formerly pushing math models at OpenAI. As Jia was participating in the code-model BigCode collaboration with some Hugging Face folks, came the idea to collaborate and explore how well code models could be used for formal mathematics. For context, olympiad math problems are extremely hard and the core of the issue is in the battle plan you draft to tackle each problem. A first focus of Numina was thus on creating high quality instruction Chain-of-Thought (CoT) data for competition-level mathematics. This CoT data has already been used to train models like DeepSeek Math, but is very rarely released so this dataset became an unvaluated ressource to tackle the challenges. BigCode's lead Leandro put Jia in touch with the team that trained the Zephyr models at Hugging Face, namely, Lewis, Ed, Costa and Kashif with additional help from Roman and Ben and the goal became to have a go at training some strong models on the math and code data to tackle the first progress prize of AIMO. And the trainings started: Jia being an olympiad coach, was intimately familiar with the difficulty level of these competitions and able to curate an very strong internal validation set to enable model selection (Kaggle submissions are blind). While iterating on dataset construction, Lewis and Ed from Hugging Face focused on training the models and building the inference pipeline for the Kaggle submissions. As often in competition it was an intense journey with Eureka and Aha moments pushing everyone further. Lewis told me about a couple of them which totally blow my mind. A tech report is coming so this is just some "along the way" nuggets that will be soon gathered in a much more comprehensive recipe and report. Learning to code: The submission of the team relied on self-consistency decoding (aka majority voting) to generate N candidates per problem and pick the most common solution. But initial models trained on the Numina data only scored around 13/50... they needed a better approach. They then saw the MuMath-Code paper (arxiv.org/abs/2405.07551) which showed you can combine CoT data with code data to get strong models. Jia was able to generate great code execution data from GPT-4 to enable the training of the initial models and get to impressive boost in performance. Taming the variance: Another Ahah moment came at some point when a Kaggle member shared a notebook showing how DeepSeek models worked super well with code execution (the model breaks down the problem into steps and each step is run in Python to reason about the next one). However, when the team tried this notebook they found this method had huge variance (the scores on Kaggle varied from 16/50 to 23/50). When meeting in Paris for a hackathon to improve this issue (like the HF team often does) Ed had the idea to frame the majority voting as a "tree of thoughts" where you'd progressively grow and prune a tree of candidate solutions (arxiv.org/abs/2305.10601). This had an impressive impact on the variance and enabled them to be much more confident in their submissions (which showed in how the model ended up performing extremely well on the test set versus the validation set) Overcoming compute constraints: the Kaggle submissions had to run on 2xT4s in under 9h which is really hard because FA2 doesn't work and you can't use bfloat16 either. The team explored quantization methods like AWQ and GPTQ, finding that 8-bit quantization of a 7B model with GPTQ was best Looking at the data: a large part of the focus was also on checking the GPT-4 datasets for quality (and fixing them) as they quickly discovered that GPT-4 was prone to hallucinations and failing to correctly interpret the code output. Fixing data issues in the final week led to a significant boost in performance. Final push: The result were really amazing and the model climbed to the 1 place. And even more, while tying up for first place on the public, validation leaderboard (28 solved challenges versus 27 for the second place), it really shined when tested on the private, test leaderboard where it took a wide margin solving 29 challenges versus 22 for the second team. As Terence Tao himself set it up, this is "higher than expected" Maybe what's even more impressive about this competition, beside the level of math these models are already capable of is how ressource contraint the participants were actually, having to run inference in a short amont of time on T4 which only let us imagine how powerful these models will become in the coming months. Time seem to be ripe for GenAI to have some impact in science and it's probably one of the most exciting thing AI will bring us in the coming 1-2 year. Accelerating human development and tackling all the real world problems science is able to tackle.

@_lewtun Nice gap to second place! ;)

After 3 months of hard work, I'm heaps excited to share that our team won the first progress prize of the AI Math Olympiad 🥇! kaggle.com/competitions/a… This challenge involved fine-tuning open LLMs to solve 50 difficult math problems spanning geometry to number theory 🤓 Our Numina Math 7B model managed to solve 29 of these problems, making it the best-in-class math LLM 💪 Stay tuned - we will release the model, dataset, and methodology very soon!

A cool thread on bees and scientific integrity

but they were, all of them, deceived, for another Ring was made. but it was an official act so there was not much they could do about it ¯\_(ツ)_/¯