Is Muon as good as they say? We looked beyond training speed and found a hidden cost: Muon loses the simplicity bias of older optimizers like gradient descent — and this matters for generalization. Led by Sara Dragutinović and advised by Rajesh Ranganath arxiv.org/abs/2603.00742

Show Post

Connect at the ELLIS Sites ICML Fest...in Lausanne 🇨🇭 on 29 June at @epfl-ai-center.bsky.social. Meet the local AI/ML community and engage with the broader ELLIS network with the ELLIS Unit Lausanne. 🔗 More info → https://bit.ly/4w5XkmO #ICML2026

Show Post

Turns out, even after being in this gig for so long*, you can still be dumbfounded and deeply upset by a paper rejection. Great, I feel young (and mad) again. *Kidding, it's only been fourteen yea... crap, I'm a dinosaur.

Show Post

Near the end of 2025, TMLR got 500 submissions every ~60 days. Most recently: it took 19 days. This is unsustainable for a volunteer team of reviewers and action editors. TMLR is implementing authorship quotas to help relieve the significant pressure on their time and energy.

Show Post

TMLR has been facing an significant uptick in the number of submissions since the start of 2026. This is placing an extreme burden on our amazing team of reviewers & action editors. To ease this burden, TMLR will be implementing submission quotas, effective July 1. 1/n medium.com/@TmlrOrg/ann...

Show Post