Alex Hägele

@haeggee

PhD Student in ML @ICepfl MLO. MSc/BSc from @ETH_en. Previously: Fellow @AnthropicAI, Student Researcher @Apple MLR.

Lausanne, Switzerland

haeggee.github.io

Joined January 2020

742 Following

1,519 Followers

Replies Reposts Media

Pinned
Alex Hägele
@haeggee
Jun 15
New research from our MLO Lab @EPFL: Improving Neural Network Training by Decoupling the Magnitude and Direction of Weight Vectors. Magnitude-Direction Decoupling (MD): a simple optimizer tweak, and what we (currently) think is the right way to train efficiently at scale. 🧵
12
Alex Hägele
@haeggee
Oct 27
nice short writeup of properties of scale invariance in neural nets. the reaction when presenting these (seemingly benign) facts is usually quite some surprise! :)
Yusuke Miyashita
@miyashita_03
Oct 26
Article
Understanding MuonH(Hyperball) through Weight Norm Geometry
TL;DR When a network is scale invariant, multiplying its weights by a positive constant leaves its output and loss unchanged. Its gradient is perpendicular to the weights. Scaling the weights by "c"
2
Alex Hägele
@haeggee
Oct 14
Apertus 1.5 tech report (134 pages!) finally out:
Lazarz
@Laz4rz
Oct 14
134 pages from ETH and EPFL on training a multimodal reasoning model I’ve had a lot of fun putting off various fires there, given the fully open source and data requirements, you won’t find a better report, especially a multimodal one
1
Alex Hägele
@haeggee
Aug 20
personal update: a few weeks ago I joined DeepMind as a student researcher in the paris office, hosted by the great @m_e_sander in the Frontier Gemma team :)
7
Alex Hägele
@haeggee
Jul 7
Come join our cracked team of researchers and engineers to work on pretraining safe models from scratch!
Bob West
@cervisiarius
Jul 6
🚨 3 AI Research Engineer positions @EPFL_en on safety pretraining & alignment for LLMs Co-hosted by #SwissAI #MLO #dlab ➡️ Will contribute directly to #Apertus—one of the world's largest fully-open LLM efforts, trained on 10k+ GPUs 🖥️ 👉 Info & app: dlab.epfl.ch/2026-07-06-saf…

Alex Hägele

Alex Hägele’s posts