Profile avatar
jonkhler.argmin.xyz
web @ https://argmin.xyz interests: machine learning, ai4science, algorithms, coding member of technical staff @ https://cusp.ai past @ MSR, DeepMind, MPI-IS home @ Heimbach (Gilserberg), Berlin, Europe born @ 353 ppm block toxicity he/him
133 posts 2,678 followers 288 following
Regular Contributor
Active Commenter

The blog post of the late Felix Hill is powerful. Stress for AI researchers today is real. I did not know Felix Hill and I am sorry for those who did. This story is perhaps a reminder for students, postdocs, founders and researchers to take care of their well being. medium.com/@felixhill/2...

Unreal landscapes ❤️

🤔🤔

@imtd.bsky.social convinced me to turn my homepage into a blog. Not much content yet... Let's see if I stick to it: argmin.xyz.

And today, happy Hanukkah 🕎❤️

Fun challenge that is apparently hard for LLMs (at least all those I could get my hands on): solving carefully drafted anagrams! New ARC challenge? :D PS: credits go to family Lehmann from Kassel who invent those ridiculously hard Xmas puzzles every year!

Merry Christmas! ❤️🌲

Segment radixsort, prefixsums and pointwise transforms are just OPed on GPUs. The amount of throughput you can get on a single cheap device is ridiculous! And you can implement so much crazy stuff with just that. ❤️

hello bluesky! we have a new preprint on solvation free energies: tl;dr: We define an interpolating density by its sampling process, and learn the corresponding equilibrium potential with score matching. arxiv.org/abs/2410.15815 with @francois.fleuret.org and @tbereau.bsky.social (1/n)

Awesome work! Amazing to see this out :)

“Sorry”

I’ll be at NeurIPS next week –happy to connect! Hit me up if you’d like to meet up and chat :) (tabular deep learning; in-context learning; SSL; AI in EU; politics; where to get good coffee/food; your favourite Parks and Rec moment; …)

lets do #rust this year :) #AdventOfCode adventofcode.com/2024/day/1

Every two days I get this "invalid handle" label for my custom handle @jonkhler.argmin.xyz (using the DNS TXT config). This seems to be a bug - whenever I reset it, it's fine for another bunch of days. Anyone else experiencing this? If so, what's the cause/fix? #bluesky @jay.bsky.team

I wrote a summary of the main ingredients of the neat proof by Hugo Lavenant that diffusion models do not generally define optimal transport. github.com/mathematical...

The link between diffusion models and optimal transport is still a bit of an enigma to me. One thing that's clear: different diffusion models trained on similar datasets tend to recover similar mappings. If these are generally not OT, in what sense are they optimal instead?

I find strange that engineering the gradient radically to learn better is not a whole field.

See who has blocked you 🫣 and added you to a list on Bluesky. clearsky.app

cuda level programming is fun - but also makes me curse a lot :D