Trustworthy ML Initiative (TrustML)
@trustworthy_ml
Latest research in Trustworthy ML. Organizers: @JaydeepBorkar @sbmisi @hima_lakkaraju @sarahookr Sarah Tan @chhaviyadav_ @_cagarwal @m_lemanczyk @HaohanWang
ID:1262375165490540549
https://www.trustworthyml.org 18-05-2020 13:31:24
1,7K Tweets
5,9K Followers
64 Following
Follow People
Thanks to ACM CCS 2023 for a Distinguished Paper Award for our work on LM Security. We show how an adversary can steal the decoding algorithm of Language Models through API access at low cost. Kudos to Ali Naseh Kalpesh Krishna Mohit Iyyer. #CCS23 Manning College of Information & Computer Sciences people.cs.umass.edu/~amir/papers/C…
I'm recruiting PhD students Allen School UW NLP (bdata.uw.edu). Focus areas include Human-AI collaboration, language agents, LLM safety & applications to mental health, social sciences, education.
Apply here: cs.washington.edu/academics/phd/…
UW Data Science
@UW_iSchool
HCI & Design at UW
✨Excited to share ACM CCS 2023 about our work on unraveling the connections between Differential Privacy and Certified Robustness in Federated Learning against poisoning attacks!🛡️🤖
🗓️ Join our talk this afternoon. Happy to discuss if you are around!
Paper: arxiv.org/abs/2209.04030
Looking forward to SaTML Conference 2024 at University of Toronto ! Feel free to reach out if you have any questions about the conference.
Dates: April 9-11
Location: University of Toronto downtown campus
just one week left of the NeurIPS Unlearning Challenge! It's been a nerve-wracking three months, and we're excited to see what the final submissions bring. 🎉 #NeurIPS2023 #UnlearningChallenge kaggle.com/competitions/n…
Are you interested in attending SaTML Conference in Toronto (April 9-11, 2024) but lacking funding to do so?
We will use funds from our sponsors to support student travel to the conference. Please apply here by December 20 to receive full consideration:
docs.google.com/forms/d/e/1FAI…
Differential privacy is a hammer, but not every privacy problem is a nail. 🔨
We introduce 'Gaussian Membership Inference Privacy' in our #NeurIPS2023 paper with Martin Pawelczyk and Gjergji Kasneci.
A thread 🧵👇 [1/n]
Fun new work led by the awesome Qingru Zhang:
Mechanistic interpretability can improve instruction-following (>20% boost for LLaMA) with:
- 👈Better user control of prompts
- 💨No extra inference cost
- 🤏Only a handful of learned parameters
arxiv.org/abs/2311.02262