Lukas Heinrich

Lukas Heinrich

@lukasheinrich.com

High Energy Physics/Machine Learning/Data Science Prof @tum.de www.lukasheinrich.com

1.1K followers935 following76 posts

Lukas Heinrich reposted

Tom McCoy
@rtommccoy.bsky.social

🤖🧠NEW PAPER🧠🤖 (The result of an 8-year project!) LLMs seem very different from symbolic systems. Yet LLMs excel in symbolic domains (e.g., language/code/math). How do they do it? Our finding: LLM representations have implicit symbolic structure! Link in thread ⬇️ 1/n

Lukas Heinrich
@lukasheinrich.com

Great place and opportunity

ML for Science@ml4science.bsky.social

We are recruiting several Early Career Research Group Leaders for Machine Learning in Science (m/f/d; E14 TV-L, 100%) in Tübingen! Open to researchers directly after their PhD or with postdoctoral experience. Application deadline: July 15, 2026. More info: uni-tuebingen.de/en/128980#c2... 1/2

Lukas Heinrich
@lukasheinrich.com

this one of these papers that was 100% worth really digesting deeply.. it's shaped a lot how I think about / understand stats in HEP. Congrats!

Kyle Cranmer@kylecranmer.bsky.social

I have great memories of this project, which developed from a narrowly scoped question to something much more impactful over more than a year. It’s interesting how terms like “Asimov” and “signal strength” and “mu” now permeate the field.

Lukas Heinrich reposted

Annalena Kofler
@annalenakofler.bsky.social

The new gravitational wave data is public, this is so exciting! Analyzing so many events is becoming increasingly expensive. That's why we have used DINGO for the first time to support the official LVK analysis. After training the neural networks, you can analyze suitable events over night! 🥳🤩

EGO and the Virgo Collaboration@egovirgo.bsky.social

The @ligo.org – Virgo – KAGRA Collaboration published today a new catalog of gravitational wave events that adds 161 events to the collection, bringing the total number of gravitational wave signals detected to date to 390 🤯

Lukas Heinrich reposted

Timothy Gowers
@wtgowers.bsky.social

OpenAI's claim that this is a central conjecture in discrete geometry is not an exaggeration. This will I think be looked back on as the first time that AI solved a major mathematics problem (defined as a problem that all experts in some subfield had thought about). openai.com/index/model-...

openai.comAn OpenAI model has disproved a central conjecture in discrete geometryAn OpenAI model solved the 80-year-old unit distance problem, disproving a major conjecture in discrete geometry and marking a milestone in AI-driven mathematics.
Lukas Heinrich
@lukasheinrich.com

I'm very excited about this new paper, which kicks of our LEGO ERC project we started earlier this year towards general-purpose AI surrogates for simulating radiation-matter interaction. Instead of training AI surrogates for specific detectors or material distribution ...

View full thread
Lukas Heinrich
@lukasheinrich.com

but with the benefit of tractable likelihoods, differentiability (and perhaps acceleration). With this paper, we give a first reference implementation that hopefully establishing a new direction for improvement for the next few years

Lukas Heinrich
@lukasheinrich.com

I'm very excited about this new paper, which kicks of our LEGO ERC project we started earlier this year towards general-purpose AI surrogates for simulating radiation-matter interaction. Instead of training AI surrogates for specific detectors or material distribution ...

Lukas Heinrich
@lukasheinrich.com

can we train one that works for many detectors our of the box? The idea is: try to learn a composable building block (hence LEGO/BRICKS :) ) of next-particle prediction that distills the collective effects of particle interacting with matter that can be iterated just like mechanistic simulators

Lukas Heinrich
@lukasheinrich.com

I always enjoy coming to Nikhef - and happy that my finishing PhD Malin Horstmann will start her Postdoctoral there - what a great place!

Tristan du Pree@tristandupree.bsky.social

Inspiring colloquium at @nikhef.bsky.social by @lukasheinrich.com on AI in Particle Physics

Lukas Heinrich reposted

Johannes Köster
@johanneskoester.bsky.social

Today, the #SnakemakeHackathon2026 at the TU Munich ended with the release of Snakemake 9.17! I want to thank all participants and my co-organizers! You have been incredibly dedicated and we improved a ton of things throughout the ecosystem.

Lukas Heinrich
@lukasheinrich.com

Very cool talk by @annalenakofler.bsky.social - check it out !

Annalena Kofler@annalenakofler.bsky.social

This is the first ever detected gravitational wave originating from a black hole merger observed in 2015. In the GIF: • Grey = the actual noisy detector signal • Color = a best-fit simulation • Left = the underlying physics: two simulated black holes spiraling inwards and merging into one

Lukas Heinrich reposted

Christian Grefe
@cgrefe.bsky.social

Vor 10 Jahren haben wir mit einer handvoll Leuten yhep.desy.de gegründet. Heute war der Hörsaal voll gepackt mit mehr als 100 early-career Physiker*innen die über das #WissZeitVG und Nachhaltigkeit in der Wissenschaft informiert wurden und sich vernetzen konnten. #IchBinHanna

Lukas Heinrich reposted

Transactions on Machine Learning Research
@tmlrorg.bsky.social

TMLR's 2025 annual report! The word of the year is growth: whether it's our number of submissions, published papers, or reviewers, pretty much everything has increased ~50%. EXCEPT: our review period, which is still shorter than NeurICMLR (91 vs 126 days) Read on for more! 1/n

Lukas Heinrich reposted

ESO
@eso.org

1/ Happy International Day of #WomenAndGirlsInScience! To mark the occasion, women from across the @eiroforum.org organisations, including ESO, shared one piece of advice that helped them in their professional journeys 👇 Discover more: www.eiroforum.org/news/eirofor...

EIROforum@eiroforum.org

Happy International Day of Women and Girls in Science! 👩‍🔬 To celebrate, we asked women from across our member organisations to share one piece of advice they received that helped them get to where they are today. What advice has made an impact on you! Read more: www.eiroforum.org/news/eirofor...

Lukas Heinrich
@lukasheinrich.com

Roger‘s Statistics Book is a staple among particle physicists - RIP

William Barter@wjbarter.bsky.social

Profoundly sad to read news of Roger Barlow's passing. A great physicist, educator and communicator, who lived by his principles and would share his wisdom with anyone who asked. The world is now a poorer place.

Lukas Heinrich
@lukasheinrich.com

Scaling Laws in Particle Physics Data! This is a result I've been itching to share and it's finally out. One of the big open questions is how much better AI-based methods at particle colliders can still become. 1/4

View full thread
Lukas Heinrich
@lukasheinrich.com

Instead of training on ~300M jets we train on up to 10 Billion(!) jets and see not only performance improve but improve predictably with the ultimate limit seemingly still far away. Now that we have the scaling laws, we can try to start understanding them. 3/4

Lukas Heinrich
@lukasheinrich.com

Scaling Laws in Particle Physics Data! This is a result I've been itching to share and it's finally out. One of the big open questions is how much better AI-based methods at particle colliders can still become. 1/4

Lukas Heinrich
@lukasheinrich.com

This paper led by Matthias Vigl, Nicole Hartman and Nikita Pond shows indeed there is still *much* more to gain and that the gain is predictable similar to the scaling laws we've seen in language and vision. 2/4

Lukas Heinrich
@lukasheinrich.com

Very interesting work by our own @rythian47.bsky.social that I finally understood a bit better recently when he presented it at our MIAPbP workshop. Happy to see this accepted at Neurips!

Baran Hashemi@rythian47.bsky.social

We got accepted at #NeurIPS2025. I am very happy that I could merge my knowledge of Mathematics with AI to create sth new and useful for the community. ☺️ The paper: arxiv.org/abs/2505.17190 The code: github.com/Baran-phys/T...

arxiv.orgTropical Attention: Neural Algorithmic Reasoning for Combinatorial AlgorithmsDynamic programming (DP) algorithms for combinatorial optimization problems work with taking maximization, minimization, and classical addition in their recursion algorithms. The associated value func...

Lukas Heinrich reposted

Keenan Crane
@keenancrane.bsky.social

I got tired of mashing together tools to write long threads with 𝐫𝐢𝐜𝐡 𝑓𝑜𝑟𝑚𝑎𝑡𝑡𝑖𝑛𝑔 and ℳα†ℏ—so I wrote La𝑇𝑤𝑒𝑒𝑡! It converts Markdown and LaTeX to Unicode that can be used in “tweets”, and automatically splits long threads. Try it out! keenancrane.github.io/LaTweet/

Lukas Heinrich reposted

Patrick Koppenburg 🐧
@koppenburg.ch

RIP Herwig Schopper, CERN director general from 1973 to 1976. He made LEP possible, and thus also the LHC. Not so long ago I worked with him on a book about the scientific discoveries at CERN. Read the interview in the CERN courier about the lessons from LEP cerncourier.com/a/lessons-fr...

cerncourier.comLessons from LEP – CERN CourierThe Large Electron Positron collider changed particle physics forever. As the field eyes up the next major collider, former CERN Director-General Herwig Schopper describes what it took to make LEP…
Lukas Heinrich
@lukasheinrich.com

Big grant news today! I feel very lucky and honored for this opportunity from @erc.europa.eu . We will attempt to go for a big qualitative step up in how we use AI/ML to predict how particles interact with matter. Stoked to get started on this in 2026. we will release job advertisements soon!

Technische Universität München@tum.de

Congratulations to six of our #researchers, including Lukas Heinrich, who receive prestigious ERC #StartingGrants worth up to 1.5 million euros each for projects in #informatics, #medicine, #LifeSciences & #NaturalSciences: go.tum.de/682017 👏 #ERCStG @erc.europa.eu 📷A.Eckert

go.tum.deSix ERC Starting Grants for researchers at TUMSix further researchers at TUM are to receive the prestigious ERC Starting Grants for their projects.
Lukas Heinrich
@lukasheinrich.com

2/ the behavior (e.g. is 2nd descent better 1st?) strongly depends on the details . Recently HEP has been pushing on scaling our models and understanding these types of phenomena from a deep learning theory perspective would be great! It’d be great to connect to some CS folks on this.

Lukas Heinrich reposted

Eli Weinstein
@eliweinstein.bsky.social

I'm looking for my first PhD student! We will push the frontiers of probabilistic machine learning for the molecular sciences, and study how to design new algorithms that exploit the unique properties of molecular systems to learn about the world. efzu.fa.em2.oraclecloud.com/hcmUI/Candid...

efzu.fa.em2.oraclecloud.comPhD scholarship in Machine Learning for Molecules - DTU ChemistryAdvance large scale data generation for chemical and biological AI in a 3 year PhD. Work on the frontier of active learning, and develop novel probabilistic machine learning techniques for experiment ...