Zachary Charles (@matharycharles) 's Twitter Profile
Zachary Charles

@matharycharles

distributed machine learning @ google | sometimes mathematician

ID: 848514751

linkhttp://zachcharles.com calendar_today27-09-2012 02:20:29

508 Tweet

996 Followers

312 Following

Gal Mishne ๐Ÿ’”๐Ÿ‡ฎ๐Ÿ‡ฑ (@gmishne) 's Twitter Profile Photo

Serious issues with AC paper bidding NeurIPS Conference Paper matching ignored my excluded papers, gave me mostly papers on topics I have little to no expertise in and we can only bid on limited subset of these mostly irrelevant papers. Friend reported having the same issue #NeurIPS25

Michael Dennis (@michaeld1729) 's Twitter Profile Photo

Anyone else finding the NeurIPS reviewer biding impossible this year? The affinity score is a poor proxy for my ability to review a paper, and having only the first 90 by that metric leaves papers mostly requiring significant background I am not particularly close to having.

Zachary Charles (@matharycharles) 's Twitter Profile Photo

Why are there barely any papers for me to bid on for NeurIPS? They say to bid affirmatively on 30-40 papers but I'm only being shown about that many (and most of them are not in areas I care about).

Zachary Charles (@matharycharles) 's Twitter Profile Photo

I've heard a lot of people talking about. How TMLR alleviates many of the problems of conference reviewing. I agree - but it needs more reviewers. If you're feeling frustrated about ML conferences please sign up to be a reviewer!

Zachary Charles (@matharycharles) 's Twitter Profile Photo

We showed that an enhanced version of Lookahead worked great for LLM pre training. See arxiv.org/abs/2503.09799 - this super Lookahead method is actually just DiLoCo without the parallelism.

Neal Parikh (@npparikh) 's Twitter Profile Photo

If your reaction to this is to say, who cares, whoโ€™s heard of Gรถttingen anyway? โ€” exactly, and you are the bad guy in the story. Gรถttingen was the epicenter of math, with a cast of characters following Gauss too long to tweet. Then the Nazis came and it was gone.

Zachary Charles (@matharycharles) 's Twitter Profile Photo

Last math quote of the day. Goro Shimura remembering Yutaka Taniyama, in the form of praise that one can only hope to aspire to.

Last math quote of the day. Goro Shimura remembering Yutaka Taniyama, in the form of praise that one can only hope to aspire to.
Zachary Charles (@matharycharles) 's Twitter Profile Photo

DiLoCo and related methods for communication efficient learning continues to lead to interesting optimization work for LLMs! Excited to read this when I'm back from vacation.