Rosanne Liu

@savvyRL

Cofounded & running @ml_collective. Weekly talks at Deep Learning Classics & Trends (DLCT). Research at Google DeepMind. DEI/DIA Chair of ICLR & NeurIPS.

San Francisco, CA

rosanneliu.com

Replies Reposts Media

Pinned
Rosanne Liu
@savvyRL
Jul 10, 2025
The opportunity gap in AI is more striking than ever. We talk way too much about those receiving $100M or whatever for their jobs, but not enough those asking for <$1k to present their work. For 3rd year in a row, @ml_collective is raising funds to support @DeepIndaba attendees.
22
Rosanne Liu
@savvyRL
Sep 2
Stop blowing up Looped Transformers — it's not magical, it doesn't add recurrence to the (still and always) feedforward network, it was done years ago (Dehghani et al., 2019). All it does is making your model deeper, the same as if you added more layers to begin with.
13
Rosanne Liu
@savvyRL
Aug 20
100%! The title we secretly want to publish it under is "Recirculation. OR: CoT is a cop out."
Lucas Gray-Joy
@techdreamergc
Aug 20
Replying to @mc_mozer
Interesting distinction from chain-of-thought. CoT is the default for reasoning gains, but state tracking feels more fundamental. If feedforward depth limits state updates, inference-time recurrence is a natural fix.
4
Rosanne Liu
@savvyRL
Aug 19
18 years into paper publishing and I still sheepishly search for our new paper title to see if it's gotten picked up in the gap between arxiv posted and twitter announced
2
Rosanne Liu
@savvyRL
Jul 27
On a two week vacation aka full time childcaring with no pay no appreciation and 24h on call where your coworker is kinda a jerk but super duper cute
6