maybe it's a skill issue of mine, but I still really really hate LLMs for long-form writing. I've tried many different ways: asking for a v0 and fixing myself. Writing a v0 and asking to improve, doing it piece by piece, one-shot, skills md file. everything still feels like garbage and no taste.

scaling laws appeared as "learning curve models" long ago in the literature (90's), and finally people are going back to it. I love the chinchilla paper but that's not the whole story/method arxiv.org/abs/2509.19189

Functional Scaling Laws in Kernel Regression: Loss Dynamics and Learning Rate Schedules

Scaling laws have emerged as a unifying lens for understanding and guiding the training of large language models (LLMs). However, existing studies predominantly focus on the final-step loss, leaving o...

arxiv.org

In the same way that mathematicians argue that their field has never been about the answers, but about the questions and their proofs, computer scientists should repeat Leslie Lamport's mantra: coding is not programming.

I'm trying to write about the history of scaling laws, and my go-to reference in the ML community is [1]. If anyone has good suggestions in asymptotic stats, I'm curious to read and help make the connections. [1] proceedings.neurips.cc/paper_files/...

I've finally deleted my twitter account, but as much as I love bksy's idea, it doesn't seem to be a good replacement. In terms of keeping up with science, linkedin and reddit have unbelievably proven more effective for me. Is there anything I'm missing here? A feed to follow, a better way to use it?

I can only imagine how crazy it must be to be a PhD student submitting to ML conferences now. The process has always been noisy, but at this point it's selecting for either obfuscation or shallow ideas. You either intimidate the reviewer, or you write a blog post in latex.

We're excited to present our latest article in Nature Machine Intelligence - Boosting the predictive power of protein representations with a corpus of text annotations. Link: www.nature.com/articles/s42... [1/4]

the goat of brazilian music w/ the best of (current) american music www.youtube.com/watch?v=jFUh...

Milton Nascimento & esperanza spalding: Tiny Desk (Home) Concert

YouTube video by NPR Music

youtube.com

I loved this new preprint by Lourie/Hu/ @kyunghyuncho.bsky.social . If you really wanna convince someone youre training a foundation model, or proposing better methodology, loss scaling laws aren't enough. It has to be tied w/ downstream performance. it shouldn't be vibes arxiv.org/abs/2507.00885

Scaling Laws Are Unreliable for Downstream Tasks: A Reality Check

Downstream scaling laws aim to predict task performance at larger scales from pretraining losses at smaller scales. Whether this prediction should be possible is unclear: some works demonstrate that t...

arxiv.org

I'm very excited about our new work: SciGym. How can we scale scientific agents' evaluation? TLDR; Systems biologists have spent decades encoding biochemical networks (metabolic pathways, gene regulation, etc.) into machine-runnable systems. We can use these as "dry labs" to test AI agents!

I wish we had an ML equivalent of SOSA (Symposium On Simplicity in Algorithms). "simpler algorithms manifest a better understanding of the problem at hand; they are more likely to be implemented and trusted by practitioners; they are more easily taught" www.siam.org/conferences-....

Please check out our new approach to modeling somatic mutation signatures. DAMUTA has independent Damage and Misrepair signatures whose activities are more interpretable and more predictive of DNA repair defects, than COSMIC SBS signatures 🧬🖥️🧪 www.biorxiv.org/content/10.1...

Damage and Misrepair Signatures: Compact Representations of Pan-cancer Mutational Processes

Mutational signatures of single-base substitutions (SBSs) characterize somatic mutation processes which contribute to cancer development and progression. However, current mutational signatures do not ...

biorxiv.org

I haven't been up to date with the model collapse literature, but it's crazy the amount of papers that consider the case where people only reuse data from the model distribution. This never happens, there's always some human curation or conditioning that yields some type of "real-world, new, data".

This is my favourite "graph paper" of the last 1 or 2 years. We also need to start including non-NN baselines, e.g. fingerprints+catboost ---if the goal is real-world impact and not getting it published asap. I also recommend following @wpwalters.bsky.social's blog. arxiv.org/abs/2502.14546

Position: Graph Learning Will Lose Relevance Due To Poor Benchmarks

While machine learning on graphs has demonstrated promise in drug design and molecular property prediction, significant benchmarking challenges hinder its further progress and relevance. Current bench...

arxiv.org

Unbelievable news. Pancreatic is one of the deadliest cancers. New paper shows personalized mRNA vaccines can induce durable T cells that attack pancreatic cancer, with 75% of patients cancer free at three years—far, far better than standard of care. www.nature.com/articles/s41...

After 6+ months in the making and over a year of GPU compute, we're excited to release the "Ultra-Scale Playbook": hf.co/spaces/nanot... A book to learn all about 5D parallelism, ZeRO, CUDA kernels, how/why overlap compute & coms with theory, motivation, interactive plots and 4000+ experiments!

The Ultra-Scale Playbook - a Hugging Face Space by nanotron

The ultimate guide to training LLM on large GPU Clusters

hf.co

I've always hated the "reasoning models" for code assistance since I think the most useful application of LLMs is really writing the boring helper functions and letting us focus on the hard work. However, I found o3 to be particularly useful when debugging ML code, e.g., 1/2

The whole DeepSeek-R1 thing just highlights computer science's main feature: you can do A LOT with a small team and some (limited) resources. This is how we've been able to scale innovation and why free software is important.

This is an amazing resource (of resources) for machine learners

Pat Walters@wpwalters.bsky.social · 2y ago

Machine Learning in Drug Discovery Resources page updated for 2025. github.com/PatWalters/r...

Sad after #AISTATS2025 and #ICLR2025 notifications? As we say in Italy, when a door closes, a bigger one opens ;) If you have a fantastic paper on #uncertainty #AI #ML #causality #statML #probabilisticmodels #reasoning #impreciseprobabilities etc, consider submitting to #UAI2025 🇧🇷 deadline 10 Feb 💥

uai2026@auai.org · 2y ago

The 41st Conference on #Uncertainty in #AI will be held in Rio de Janeiro 🇧🇷, July 21-25! The CfP is out 👉 www.auai.org/uai2025/call... 🚨 Feb 10: Paper submission 🗣️ Apr 3-10: rebuttal period 🎉/💀 May 6: Author notification #UAI2025 #ML #stats #learning #reasoning #uncertainty

Slides of my presentation "Mathematical Foundations of Graph Foundation Models" yesterday at the AMS Session of the #JMM2025. The accompanying paper is coming soon. www.cs.purdue.edu/homes/ribeir...

Learning Rust ~properly~ during my break and wow -- absolutely worth it! While we're all chasing GPU optimization, there's something magical about crafting efficient CPU-based apps. Clean and fast data processing can change our lives ;)

My model is that these things are extremely helpful above some skill bar and extremely harmful below some skill bar

very cool observations about using smiles/graphs vs fingerprints. TDLR; fingerprints only capture certain properties marginally, and their combinations can often give rise to something new/different. www.deepmedchem.com/articles/wha...

Presenting our poster at NeurIPS! Please come chat about estimating causal effects from user/patient-reported experiences: Thursday, 11AM, West Ballroom A-D #5110.

Rahul G. Krishnan@rahulgk.bsky.social · 2y ago

@nikitadhawan.bsky.social developed NATURAL (www.cs.toronto.edu/~nikita/natu...) with @cottascience.bsky.social , Karen & @cmaddis.bsky.social. Its an end-to-end pipeline that starts from raw-text data and ends with a causal (**) effect associated with an intervention. (**) conditions apply 🧵(6/7)

If you're interested in {causality, language, healthcare}, stop by! Thursday 11am - 2pm West Ballroom A-D #5110

Rahul G. Krishnan@rahulgk.bsky.social · 2y ago

@nikitadhawan.bsky.social developed NATURAL (www.cs.toronto.edu/~nikita/natu...) with @cottascience.bsky.social , Karen & @cmaddis.bsky.social. Its an end-to-end pipeline that starts from raw-text data and ends with a causal (**) effect associated with an intervention. (**) conditions apply 🧵(6/7)

I'm told this is a more intellectual version of ML Twitter :). I have a question... What papers have made good *theoretical* advances towards graph foundation models? Jan 8th 1-2pm I am giving a talk at the Joint Mathematics Meeting on the topic meetings.ams.org/math/jmm2025...

b] ~Billions of dollars each year are spent on trials to assess interventions. Can we use crowdsourced data to know which intervention is likely to work ahead of time? Doing so requires answering a causal question! But the data to answer this question is locked in unstructured text. 🧵(5/7)

What are the most interesting datasets and benchmark-related work for ML in drug discovery at NeurIPS? We’ll be at the conference doing short interviews with researchers and handing out some Polaris merch! Here’s who we have on the shortlist. 🧵

1/6 We're excited to share our #NeurIPS2024 paper: Probabilistic Graph Rewiring via Virtual Nodes! It addresses key challenges in GNNs, such as over-squashing and under-reaching, while reducing reliance on heuristic rewiring. w/ Chendi Qian, @christophermorris.bsky.social @mniepert.bsky.social 🧵