A must-read for supervisors and managers👇 Sexual harassment is far more common than discussed because victims often experience fear, not anger, and may freeze rather than confront.

Accepted to Conference on Language Modeling! Check out our new work on fairness implications of low-rank adaptation: arxiv.org/abs/2405.17512

We are excited to announce that 58 excellent papers will be presented at the ICML Conference TF2M Workshop. List of accepted papers: openreview.net/group?id=ICML.… You can find the detailed schedule on our website (and below 👇): sites.google.com/view/tf2m

Excited to share the program and list of accepted papers for our ICML Conference workshop Theoretical Foundations of Foundation Models: sites.google.com/view/tf2m/sche… Looking forward to discussing efficiency, responsibility, and principled foundations of foundation models in Vienna soon!

I'll be ICML Conference next week. Text me if you want to chat about trustworthy & efficient AI or data valuation 🙌 -Sat: Check out our workshops Theoretical Foundations of Foundation Models & Workshop on Data-centric Machine Learning Research -Thu 1:30-3 pm: arxiv.org/pdf/2405.02341 w/ Wei-Ning Chen Peter Kairouz Albert No Sewoong Oh and Zheng Xu👇

Join us for the Workshop on Theoretical Foundations of Foundation Models ICML Conference tomorrow! We have a fantastic list of invited speakers: - Ananda Theertha Suresh - Dan Alistarh - Jason Lee - Kamalika Chaudhuri - Yuandong Tian 1/3

Join us at TF2M workshop at Strauss 2 with an amazing lineup of speakers and panelists!

How time flies! It's the last day of #ICML2024 already. I'll be giving another talk on Lottery Ticket Adaptation at the WANT workshop today at 12:00pm. It's been a fun week (3 talks, 7 posters) and I'm looking forward to meeting more people today and chatting about research.

It was a really fun ICML Conference workshop with amazing speakers, panelists, poster presentations, and a highly engaged audience! Theoretical Foundations of Foundation Models

It was very fun participating in the Theoretical Foundations of Foundation Models. A big thanks to the organizers Berivan Isik Ziteng Sun Banghua Zhu Enric Boix-Adsera Merve Gurel Bo Li Ahmad Beirami Sanmi Koyejo!

.Stanford University finished with 39 medals at the Olympics. That puts it behind only five countries, with Italy just ahead at 40. Fear the tree.

I have joined Google AI as a research scientist. I will continue to work on efficient and trustworthy AI, LLMs, safety, and privacy. Stay tuned for updates 👀

Excited to share Lottery Ticket Adaptation (LoTA)! We propose a sparse adaptation method that finetunes only a sparse subset of the weights. LoTA mitigates catastrophic forgetting and enables model merging by breaking the destructive interference between tasks. 🧵👇

I promise I've done other things since 2018!!! That said, neat looking paper, and cool to see my most recent work with Dan Biderman (first author) connected to my most notorious work. Just wish Databricks Slack wasn't so active every time "lottery ticket" is in a paper title.

It makes sense that OpenAI Finetuning uses LoRA because the resulting adapters are easy to store and fast to deploy. But LoRA finetuning is uniquely bad at removing the alignment that OpenAI spends so much effort applying! LoTA adapters are small, but don't remove safety👀