@AurelienLucchi

Researcher in deep-learning optimization, LLM training, and reasoning 🧠. Associate professor at UniBasel. Past: EPFL, ETH Zurich (Switzerland) 🇨🇭

 Zurich
omls.dmi.unibas.ch/en/
Joined February 2013
  • Tweets364
  • Following435
  • Followers1.9K
27Media
Aurelien Lucchi retweeted
Andrea Miele@andreamiele_17h
🔥 Diffusion LMs promise what autoregressive models can’t: fixing their mistakes. But new ones often change the noise and the sampler at once. Which matters? We tested it on the hypersphere. Noise: 0.8 points. Sampler: 25. We introduce Spherical Brownian Diffusion 🌐 👇🧵
16
45
4
220
11,216
Can we build language models that rely entirely on self-conditioning? Exciting new work exploring a different paradigm for language generation! 🚀
Nadezhda Ilieva@nadjushenka_Oct 22
Can a language model be based purely on self-conditioning? 🤔 Yes! Meet LAW 👋 (a.k.a. Loopholing All the Way!) 🧵1/8
2
11
618
Another evidence of OpenAI publishing results close to unpublished material from mathematicians who used OpenAI LLMs and app. After training on copyrighted published material are they training on unpublished material? What a bad time to be a mathematician. Should be the opposite.
doomslide
@doomslideOct 8
Check out this loser! >uploads a draft to Chat on Sept. 8th >gets scooped by Chat on Oct. 6th "The result released by OpenAI (#311) on October 6, 2026... identical to a preliminary version of the present paper that was uploaded to ChatGPT on September 8, 2026" LMAO get rekt
28
48
6
360
20,069
Aurelien Lucchi retweeted
Robin Faro@robinfaro13Oct 2
🎯 Learning is most effective at the frontier of capability: problems that are too easy or too hard teach nothing. For LLM reasoners trained with GRPO this is literal: problems the model always or never solves give zero gradient. We introduce Frontier Learning👇🧵
30
68
10
1,062
103,819
Aurelien Lucchi retweeted
arXiv.org@arxivOct 1
arXiv has updated our policy on rate limiting for all submitters. This update was made to fairly distribute moderator time & support the arXiv community of staff, volunteers, readers & authors. Please read our announcement to learn more: blog.arxiv.org/2026/10/01/up…
arXiv logo in white over a whiteboard with math symbols.

ALT arXiv logo in white over a whiteboard with math symbols.

79
496
240
1,730
760,679
Aurelien Lucchi retweeted
Shashank Tripathi
@sha2nk_tSep 18
🚨 We turned European train delays into a live league table. Germany finished last. 🇩🇪 Switzerland isn't even #1. 🇨🇭 Over the last 30 days, DelayBahn tracked 900,000+ train stops across 6 countries. Welcome to Europe's Rail Leaderboard 🏆🚆 See where your country ranks 👇 delaybahn.com/rangliste Some of the results are wild 🧵 (1/6)
212
322
178
3,300
422,573
Aurelien Lucchi retweeted
Quentin Berthet@qberthetSep 9
#AISTATS 2027 Call for Paper is out! This year's edition is taking place in Montreal on May 3-6th, 2027 🇨🇦 Aymeric Dieuleveut and I are chairing virtual.aistats.org/Conferen… New this year: AI review, and more! Abstract deadline: September 29, 2026 Paper deadline: October 6, 2026
4
15
1
77
4,745
Aurelien Lucchi retweeted
Boaz Barak
@boazbaraktcsSep 8
Important points. Note also that users can opt out from use of their (de-identified) data in training. Anthropic has similar policies, though I do not remember such discussions when they announced mathematical results.
Mark Chen
@markchen90Sep 8
Two things to distinguish: Did any human or agent look at user data as part of the Navier Stokes effort? No. Do we use user feedback and de-identified data to improve ChatGPT and Codex in a holistic way? Yes. And so does every LLM company.
21
11
8
152
76,277
Anthropic: “Claude Team plan for Scientists!” Me: applies Anthropic: “We were unable to verify that you are a scientist.”
2
13
1,705
Aurelien Lucchi retweeted
Jostein Hauge
@haugejosteinAug 6
London’s Hyde Park has transformed into the Dubai desert.
448
1,551
466
15,567
4,083,359
Nice. Didn't know this was an option.
Jean Kossaifi
@JeanKossaifiJul 28
PSA for anyone preparing a NeurIPS rebuttal (or any OpenReview response). Draft in Google Docs, go to Tools > Preferences > Enable Markdown. When you’re done you can simply select the text, right-click, and “Copy as Markdown”. No more fixing formatting by hand. You’re welcome!
2
1,066
We’re hiring! 🇨🇭 Join the University of Basel as a PhD student or postdoctoral researcher in the field of LLM training. Work on: 🤖 LLM pre- and post-training 🎯 Reinforcement learning and reasoning 📈 Reliable generalization ⚙️ Large-scale experiments 📍 Basel, Switzerland
8
32
1
237
21,415
🗓 Apply by July 31, 2026 🔗 jobs.unibas.ch/offene-stelle… #LLM #MachineLearning #PhD #Postdoc #AIJobs #Switzerland
1
1
8
1,329
This is a joint position with @ilijabogunovic, so you get two advisors instead of one.
1
5
885
Aurelien Lucchi retweeted
Kimi.ai
@Kimi_MoonshotJul 27
Releasing the model weights and technical report of Kimi K3. Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window. New model architecture: 2.5x the intelligence per unit of compute, not just more params. Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale. Model weights: huggingface.co/moonshotai/Ki… Tech report: github.com/MoonshotAI/Kimi-K… Tech blog: kimi.com/blog/kimi-k3
1,544
7,191
2,353
46,006
14,747,177
Aurelien Lucchi retweeted
Mark Schmidt
@MarkSchmidtUBCJul 21
Slides from my ICML tutorial "Is numerical optimization theory irrelevant to machine learning practice in 2026?": cs.ubc.ca/~schmidtm/Document… (Updated to fix some typos, incorporate feedback, and add some things I did not have time for. Will link to the video here when posted.)
5
84
5
539
45,924
Aurelien Lucchi retweeted
Ivan Dokmanić
@ivodokmanicJul 8
(1/6) A Warp Drive for Neural PDE Solvers??? Tomorrow in Seoul, Till and Alex are presenting their floral #ICML2026 Spotlight. Don’t miss it if you care about neural surrogates for physics, foundation models, physics world models, …
2
14
69
5,739
Aurelien Lucchi retweeted
Antonio Orvieto
@orvieto_antonioJun 17
Judging optimizer gaps by looking only at language modeling with a fixed batch size is dangerous: one gets only 1/2 of the story. @orientino_ and @ruuustem_10 went beyond. Turns out that for every model, task, and data, there is always a setup where Adam > SGD. 🧵
2
15
3
74
10,626
Aurelien Lucchi retweeted
Ilija Bogunovic
@ilijabogunovicApr 20
Hiring 2 summer ML research interns at the University of Basel 🇨🇭. Research topics: RL/diffusion LLM post-training, reasoning, or LLM orchestration. Possible fully funded PhD offers to follow. I'll be at ICLR this week and happy to chat. Apply: forms.gle/TeeeNU6e7kDH3jX96
12
34
369
23,442
After a lot of hands-on time with Codex and Claude: Claude is significantly better at actually following instructions. Not even close right now.
Tibo
@thsottiauxApr 16
Codex Compute efficient ✅ Always up, never down ✅ Best at hardcore engineering ✅ Crazy good app, first to escape the terminal ✅
1
740
Load more