Profile avatar
amartyasanyal.bsky.social
Assistant Professor @Dept. Of Computer Science, University of Copenhagen, Ex Postdoc @MPI-IS, ETHZ, PhD @University of Oxford, B.Tech @CSE,IITK.
15 posts 878 followers 144 following
Regular Contributor
Conversation Starter
comment in response to post
#ICLR »Differentially Private Steering for Large Language Model Alignment« by @anmolgoel.bsky.social, Yaxi Hu, Iryna Gurevych (@igurevych.bsky.social) & Amartya Sanyal (@amartyasanyal.bsky.social) (2/🧵)
comment in response to post
Thank you!
comment in response to post
Thanks Christoph!
comment in response to post
Thank you!!
comment in response to post
@ccanonne.bsky.social : Steak holders
comment in response to post
And I think a similar argument holds for synthetic data. Synthetic data algorithms that don't provably account for privacy probably doesn't provide privacy. But there are private synthetic data generation algorithms that do like @gautamkamath.com linked above.
comment in response to post
Transformations like JL can indeed preserve privacy (arxiv.org/abs/1204.2136), while others may lead to (quantifiable) privacy degradation (arxiv.org/abs/2403.13041). The point is perhaps that augmentations, by themselves, don’t inherently guarantee an increase or decrease in privacy.
comment in response to post
Graduating with a PhD related to privacy and robustness in machine learning? Apply to this post-doc opening by @amartyasanyal.bsky.social: employment.ku.dk/faculty/?sho...
comment in response to post
Done
comment in response to post
Would love to be added as well if possible