Vinu Sankar Sadasivan (@imVinusankars) / X

Vinu Sankar Sadasivan

310 posts

Vinu Sankar Sadasivan

@imVinusankars

AI Research Scientist @Meta Superintelligence Labs | PhD from @UMDcs | Prev. @GoogleDeepMind, @MSFTResearch, @Caltech, @IITgn | sharing my views

Menlo Park, CA

vinusankars.github.io

Joined November 2011

Vinu Sankar Sadasivan
@imVinusankars
Jul 25, 2025
I am extremely delighted to announce that I will be joining @AIatMeta as an AI Research Scientist starting next week in Menlo Park! I will be working on continuing my AI safety research efforts. Exciting… I feel blessed and grateful to my advisor, family, and friends 😇
73K
Vinu Sankar Sadasivan
@imVinusankars
May 18, 2024
Thrilled to announce that I'll be starting as a Research Scientist Intern @AIatMeta next week for the summer! I'll be working with the FAIR team in Paris🇫🇷 on GenAI related security issues. If anyone is around and wants to chat about AI security/privacy, I'd love that!
16K
Vinu Sankar Sadasivan
@imVinusankars
Jul 20, 2024
Replying to @al1dao
Sure :) pls ping me, i think your inbox is not open
86K
Vinu Sankar Sadasivan
@imVinusankars
Jun 18, 2025
📢Exciting! I have successfully completed my #PhD from @ml_umd on AI Safety🥳 Grateful to my family, teachers, and friends😇 Special thanks to the most supportive advisor @FeiziSoheil and committee members! Thank you to all those who attended my defense and to my collaborators❣️
8.6K
Vinu Sankar Sadasivan
@imVinusankars
Apr 4, 2021
I am thrilled to announce that I have officially accepted an offer to join @umdcs for my PhD! 🥳🥳🥳 Grateful to my family, teachers, and friends for what I have achieved.
Vinu Sankar Sadasivan
@imVinusankars
May 14, 2024
Happy to announce📢 that our work on Fast Adversarial Attacks on LLMs *BEAST* has been accepted at #ICML24 See our code here to Jailbreak in 1 GPU-Minute💡github.com/vinusankars/BE… News🚀 1⃣BEAST to jailbreak GPT3.5 & GPT4 in black-box manner 2⃣ Improved readability for prompts
8.7K
Vinu Sankar Sadasivan
@imVinusankars
Feb 27, 2024
We are excited to share BEAST that can attack #LLMs in 1 GPU Minute! BEAST can cause #hallucination in aligned LLMs 🤯 #Jailbreak Vicuna to get 90% success in 1 GPU Minute! Boost existing #privacy attacks with our adversarial prompts We hope BEAST can accelerate S&P research
6.2K
Vinu Sankar Sadasivan
@imVinusankars
Mar 28, 2023
📢 Our #CVPR2023 work shows that controlled blurring with #random filters can poison image datasets! - No need to learn adversarial noises - Stealth x poisoning trade-off - Theoretical existance of such poisons 🤔Why? 🧵👇 CUDA: arxiv.org/abs/2303.04278
8K
Vinu Sankar Sadasivan
@imVinusankars
Jun 17, 2023
Kulkarni Fellowship will help me in advancing my research in #AI security and privacy. I'm grateful for receiving this award! Thanks to my wonderful advisor @FeiziSoheil ! Thanks to my department @umdcs @umiacs @ml_umd and the Kulkarni Foundation.
UMD Department of Computer Science
@umdcs
Jun 17, 2023
Our @UofMaryland graduate student Vinu Sankar Sadasivan has received the Kulkarni Foundation Summer Research Fellowship. Sadavisan will use the fellowship to continue his research into issues of security and privacy in artificial intelligence. Read more: go.umd.edu/Vinu
4.8K
Vinu Sankar Sadasivan
@imVinusankars
May 21, 2024
Replying to @imVinusankars
Unexpectedly met @ylecun on my first day at internship in Paris @AIatMeta ! 😇
4.2K
Vinu Sankar Sadasivan
@imVinusankars
Jan 9, 2025
I am currently on the job market looking for Research Scientist positions. If you have any opportunities related to AI safety/trustworthiness, please let me know. My PhD works revolve around AI content detection (watermarking, evasion) and AI jailbreaking (red teaming)
976
Vinu Sankar Sadasivan
@imVinusankars
Apr 21, 2023
So glad to be recognized as a notable reviewer 😇 #iclr2023
1.4K
Vinu Sankar Sadasivan
@imVinusankars
Jul 23, 2025
Can we add semantic watermarks to AI-generated images to make them more robust and interpretable to trained humans? For e.g., in the simple case of a cat image below, the images on the right are watermarked. The watermark here is the coffee mug on the stack of newspapers! 🧵1/
3.5K
Vinu Sankar Sadasivan
@imVinusankars
Jul 22, 2024
Want to know how to jailbreak language models in just one GPU minute? Come visit our poster at #ICML2024 Poster session 1 #2504. Learn how to use our fast adversarial attack BEAST to make LLMs hallucinate or break safety alignment or boost privacy attacks.
5.2K