Neel Patel

Research Scientist @GoogleResearch working on LLM agent security, inference scaling and post-training.
PhD in Theoretical CS from @USC

[Neel Patel @Neel88pat]

@nsrg_shah Congratulations Nisarg! Very well deserved :-)

[Neel Patel @Neel88pat]

@Carol_Marge_J Congrats Caroline :-)

[Neel Patel @Neel88pat]

Looking at all the tweets about amazing work in LLMs and computational linguistics, I am wondering how can someone with mathematical/theoretical CS background can contribute in the field. Any pointers/links are appreciated. Thanks :-)

[William Wang @WilliamWangNLP]

Probably the best tutorial on RL out there.

[Karthik C. S. @karthikcs_]

"ETH implies PIH"! (https://t.co/1qeyz1VqKr…)
Breakthrough work!
Hopefully this work will pave the way to proving PIH in it's original formulation (which is W[1]-hardness of gap 2-CSP).

[Ritwik Mishra @koiRitwikHai]

Canada visa refused 🙁 reasons attached

But
1) My entire family is in India. I have no history of foreign travel.
2) ACL conference is 9th to 14th july, I applied for a stay from 8th to 15th july.

What to keep in mind while applying next time? 🙇‍♂️
@aclmeeting @CitImmCanada

[Nancy Gupta @9c_gupta]

When I was in my undergrad, I used to attend the talk and session where they talked about “subtle discrimination” against women at workplaces. Back then I doubt the intensity of the matter.
#women_in _stem #lean_in #women_at_work

[Neel Patel @Neel88pat]

In our recent work: https://t.co/IbhTnz2Skw
we answer this question affirmatively for several optimization problems including additive and submodular optimization over matching, matroid and intersection of matroids.

[Neel Patel @Neel88pat]

Given a limited query excess to stochastic (combinatorial) input of the optimization problem, is it possible to select a sparse subset of the input such that you’re guaranteed to observe a good solution? (1/2)

[Neel Patel @Neel88pat]

I am also giving a talk about our work at @FAccTConference on 24th June at 12:45 pm (local Seoul time).

[Reza Shokri @rzshokri]

Can we explain models while preserving data privacy? @FAccTConference #FAccT2022 In a joint work with @Neel88pat and @YairZick, we design DP algorithms for model explanations, and study the implications of DP on the variance and usefulness of explanations. https://t.co/RxHaBX1oSM

[Nicole Wein @WeinNicole]

A very nice presentation of a cool result https://t.co/KI27GZkg6c

[Anupam Gupta @anupamg]

The list of papers to be presented at #STOC2022: https://t.co/3oD0GrVE2f

Crossing my fingers that we will be there in person!

[Jiayuan Ye @jiayuan_ye_]

Please come to our poster session if you want to learn more!

[Vladimir Sucha @VladimirSucha]

Historical step for good AI made by UNESCO. 193 UNESCO states adopted the first&only global framework for #ethics of #ArtificiallIntelligence It’s far reaching document covering all aspects people’s life. @UNESCO will be soon benchmark for ethical use of AI Congrats @gabramosp

[jack @jack]

Introducing Twitter’s first algorithmic bias bounty challenge https://t.co/hjpMv0UCqd

[Neel Patel @Neel88pat]

@leena__gh Computing expectation of a random variable:
Calculus—Hold my beer xD

[Neel Patel @Neel88pat]

This is such a beautiful article: https://t.co/cqI1dXmHSU

[Lenka Zdeborova @zdeborova]

After many talks I gave in Europe and the US this one at the most famous theoretical physics institute in India, the Tata Institute @TIFRScience https://t.co/7tfittwHNs, felt very invigorating. The 1-billion people India has a huge potential to lead science of the future!

[Neel Patel @Neel88pat]

@vkerdos Haha! I thought it was, but meh! 10 seconds feel good!

[Neel Patel @Neel88pat]

@vkerdos It was a racing incident but HE SHOULD!