Nikhil Garg
@nkgarg.bsky.social
I study algorithms/learning/data applied to democracy/markets/society. Asst. professor at Cornell Tech. https://gargnikhil.com/. Helping building personalized Bluesky research feed: https://bsky.app/profile/paper-feed.bsky.social/feed/preprintdigest
Pinned
Nikhil Garg
@nkgarg.bsky.social
· Mar 10
*Please repost* @sjgreenwood.bsky.social and I just launched a new personalized feed (*please pin*) that we hope will become a "must use" for #academicsky. The feed shows posts about papers filtered by *your* follower network. It's become my default Bluesky experience bsky.app/profile/pape...
Reposted by Nikhil Garg
This is an intriguing paper with numerous examples of using an LLM to generate formal proofs and ideas, and a nice cookbook for how to do it. Refreshingly, the paper stays away from hype and as far as possible states very clearly how the LLMs are used and the key nature of human-AI interaction.
On using LLMs for research (beyond basic prompts). Section 3.2 is specifically about catching a bug in a recent SNARG paper
arxiv.org/abs/2602.03837
arxiv.org/abs/2602.03837
Accelerating Scientific Research with Gemini: Case Studies and Common Techniques
Recent advances in large language models (LLMs) have opened new avenues for accelerating scientific research. While models are increasingly capable of assisting with routine tasks, their ability to co...
arxiv.org
February 8, 2026 at 6:14 AM
This is an intriguing paper with numerous examples of using an LLM to generate formal proofs and ideas, and a nice cookbook for how to do it. Refreshingly, the paper stays away from hype and as far as possible states very clearly how the LLMs are used and the key nature of human-AI interaction.
Reposted by Nikhil Garg
Ten math problems with proofs known the authors. Proofs are encrypted until Feb 13. For all problems, authors claim both AI-based literature searches and zero-shot attempts at proofs failed. If you want to take a crack, you have until next Friday (2/13)!
First Proof
To assess the ability of current AI systems to correctly answer research-level mathematics questions, we share a set of ten math questions which have arisen naturally in the research process of the au...
arxiv.org
February 8, 2026 at 3:59 PM
Ten math problems with proofs known the authors. Proofs are encrypted until Feb 13. For all problems, authors claim both AI-based literature searches and zero-shot attempts at proofs failed. If you want to take a crack, you have until next Friday (2/13)!
Reposted by Nikhil Garg
Our paper “Inferring fine-grained migration patterns across the United States” is now out in @natcomms.nature.com! We released a new, highly granular migration dataset. 1/9
February 5, 2026 at 5:30 PM
Our paper “Inferring fine-grained migration patterns across the United States” is now out in @natcomms.nature.com! We released a new, highly granular migration dataset. 1/9
I have been waiting for something like this to onboard onto Lean (alongside an LLM agent), and it seems like that moment is here
Never have I felt more like my job will soon by taken by AI. Statistical learning theory in Lean: concentration inequalities, Dudley's entropy integral, and local Gaussian complexity bounds.
30000 lines of code, over 1000 lemmas, formalizing Wainwright and Boucheron et al arxiv.org/abs/2602.02285
30000 lines of code, over 1000 lemmas, formalizing Wainwright and Boucheron et al arxiv.org/abs/2602.02285
February 4, 2026 at 7:20 PM
I have been waiting for something like this to onboard onto Lean (alongside an LLM agent), and it seems like that moment is here
Reposted by Nikhil Garg
🚨New WP "@Grok is this true?"
We analyze 1.6M factcheck requests on X (grok & Perplexity)
📌Usage is polarized, Grok users more likely to be Reps
📌BUT Rep posts rated as false more often—even by Grok
📌Bot agreement with factchecks is OK but not great; APIs match fact-checkers
osf.io/preprints/ps...
We analyze 1.6M factcheck requests on X (grok & Perplexity)
📌Usage is polarized, Grok users more likely to be Reps
📌BUT Rep posts rated as false more often—even by Grok
📌Bot agreement with factchecks is OK but not great; APIs match fact-checkers
osf.io/preprints/ps...
February 3, 2026 at 9:55 PM
🚨New WP "@Grok is this true?"
We analyze 1.6M factcheck requests on X (grok & Perplexity)
📌Usage is polarized, Grok users more likely to be Reps
📌BUT Rep posts rated as false more often—even by Grok
📌Bot agreement with factchecks is OK but not great; APIs match fact-checkers
osf.io/preprints/ps...
We analyze 1.6M factcheck requests on X (grok & Perplexity)
📌Usage is polarized, Grok users more likely to be Reps
📌BUT Rep posts rated as false more often—even by Grok
📌Bot agreement with factchecks is OK but not great; APIs match fact-checkers
osf.io/preprints/ps...
Reposted by Nikhil Garg
Is the only way we can create algorithms that people understand to make them trivially simple? We argue, no.
People can predict the behavior of algorithms that are arbitrarily complex, if and only if they are available, compact and aligned.
arxiv.org/abs/2601.18966
People can predict the behavior of algorithms that are arbitrarily complex, if and only if they are available, compact and aligned.
arxiv.org/abs/2601.18966
January 29, 2026 at 6:49 PM
Is the only way we can create algorithms that people understand to make them trivially simple? We argue, no.
People can predict the behavior of algorithms that are arbitrarily complex, if and only if they are available, compact and aligned.
arxiv.org/abs/2601.18966
People can predict the behavior of algorithms that are arbitrarily complex, if and only if they are available, compact and aligned.
arxiv.org/abs/2601.18966
Reposted by Nikhil Garg
🎙️ I had a great time joining the Data Skeptic podcast to talk about my work on recommender systems
If you're interested in embeddings, aligning group preferences, or music recommendations, check out the episode below 👇
open.spotify.com/episode/6IsP...
If you're interested in embeddings, aligning group preferences, or music recommendations, check out the episode below 👇
open.spotify.com/episode/6IsP...
Fairness in PCA-Based Recommenders
open.spotify.com
January 28, 2026 at 4:23 PM
🎙️ I had a great time joining the Data Skeptic podcast to talk about my work on recommender systems
If you're interested in embeddings, aligning group preferences, or music recommendations, check out the episode below 👇
open.spotify.com/episode/6IsP...
If you're interested in embeddings, aligning group preferences, or music recommendations, check out the episode below 👇
open.spotify.com/episode/6IsP...
Reposted by Nikhil Garg
There is a third-party app for this, you can find it here: bsky-follow-finder.theo.io
My impression with bluesky is that a lot of things exist but there is just no way of finding them.
@nkgarg.bsky.social
My impression with bluesky is that a lot of things exist but there is just no way of finding them.
@nkgarg.bsky.social
Bluesky Network Analyzer
Find accounts that you don't follow (yet) but are followed by lots of accounts that you do follow.
bsky-follow-finder.theo.io
January 23, 2026 at 4:14 PM
There is a third-party app for this, you can find it here: bsky-follow-finder.theo.io
My impression with bluesky is that a lot of things exist but there is just no way of finding them.
@nkgarg.bsky.social
My impression with bluesky is that a lot of things exist but there is just no way of finding them.
@nkgarg.bsky.social
This is indeed more or less exactly what happened...
Some bad decision theoretic humor, because it's my birthday
statmodeling.stat.columbia.edu/2026/01/22/a...
statmodeling.stat.columbia.edu/2026/01/22/a...
A decision theorist walks into a seminar | Statistical Modeling, Causal Inference, and Social Science
statmodeling.stat.columbia.edu
January 22, 2026 at 10:00 PM
This is indeed more or less exactly what happened...
Reposted by Nikhil Garg
Check out our new paper at #AAAI 2026! I’ll be presenting in Singapore at Saturday’s poster session (12–2pm). This is joint work with @shuvoms.bsky.social, @bergerlab.bsky.social, @emmapierson.bsky.social, and @nkgarg.bsky.social. 1/9
January 20, 2026 at 4:08 PM
Check out our new paper at #AAAI 2026! I’ll be presenting in Singapore at Saturday’s poster session (12–2pm). This is joint work with @shuvoms.bsky.social, @bergerlab.bsky.social, @emmapierson.bsky.social, and @nkgarg.bsky.social. 1/9
If you enjoy Paper Skygest, you might enjoy our new academic preprint describing it and our vision!
arxiv.org/abs/2601.04253
bsky.app/profile/pape...
arxiv.org/abs/2601.04253
bsky.app/profile/pape...
January 14, 2026 at 7:56 PM
If you enjoy Paper Skygest, you might enjoy our new academic preprint describing it and our vision!
arxiv.org/abs/2601.04253
bsky.app/profile/pape...
arxiv.org/abs/2601.04253
bsky.app/profile/pape...
Reposted by Nikhil Garg
Excited to announce that @sjgreenwood.bsky.social will be at #ATScience presenting their work on the beloved @paper-feed.bsky.social , experiments on self-hosted feeds (in collaboration with @graze.social @aendra.com) and observational analyses of social media on #atproto!
Looking forward to it! 🎉
Looking forward to it! 🎉
Exciting to see our first speaker proposals coming in for #ATScience 2026! ✨
We’d love to hear yours too, so please submit your idea at
forms.atproto.science/atscience26-... 🗳️
The call for proposals closes at the end of January - we’ll review and notify speakers on a rolling basis.
We’d love to hear yours too, so please submit your idea at
forms.atproto.science/atscience26-... 🗳️
The call for proposals closes at the end of January - we’ll review and notify speakers on a rolling basis.
ATScience2026 Call for Proposals
forms.atproto.science
January 9, 2026 at 3:01 PM
Excited to announce that @sjgreenwood.bsky.social will be at #ATScience presenting their work on the beloved @paper-feed.bsky.social , experiments on self-hosted feeds (in collaboration with @graze.social @aendra.com) and observational analyses of social media on #atproto!
Looking forward to it! 🎉
Looking forward to it! 🎉
This being the interesting bit: "All AI assistance – whether for text drafting, translation or data analysis – must be disclosed."
New Publication Ethics Policy for @informs.bsky.social. Created partially to address “emerging challenges unique to today’s publishing environment, including the use of generative artificial intelligence (AI)” pubsonline.informs.org/do/10.1287/o...
Strengthening Ethical Standards in Scholarly Publishing at INFORMS | ORMS Today
INFORMS has long been committed to advancing research of the highest quality. With over 11,000 annual submissions across 17 journals, the vitality of our publishing program reflects both the ambition ...
pubsonline.informs.org
January 6, 2026 at 5:18 PM
This being the interesting bit: "All AI assistance – whether for text drafting, translation or data analysis – must be disclosed."
Reposted by Nikhil Garg
📢 Apply by Feb 12 to join our CHI 2026 workshop, Speech AI for All, where we'll discuss inclusive speech tech for people with speech diversities. Researchers, practitioners, policymakers, & community members welcome! speechai4all.org
January 5, 2026 at 5:50 PM
📢 Apply by Feb 12 to join our CHI 2026 workshop, Speech AI for All, where we'll discuss inclusive speech tech for people with speech diversities. Researchers, practitioners, policymakers, & community members welcome! speechai4all.org
Reposted by Nikhil Garg
Friendship over with X, Bluesky is my new best friend
January 2, 2026 at 7:04 PM
Friendship over with X, Bluesky is my new best friend
Reposted by Nikhil Garg
Want to discover great posts beyond who you already follow? The For You feed, built by @spacecowboy17.bsky.social, is a personalized algorithmic feed based on your likes. It's one of our favorites for finding cool conversations across the network.
Try it here: bsky.app/profile/spac...
Try it here: bsky.app/profile/spac...
January 2, 2026 at 4:59 PM
Want to discover great posts beyond who you already follow? The For You feed, built by @spacecowboy17.bsky.social, is a personalized algorithmic feed based on your likes. It's one of our favorites for finding cool conversations across the network.
Try it here: bsky.app/profile/spac...
Try it here: bsky.app/profile/spac...
Reposted by Nikhil Garg
@eugenevinitsky.bsky.social and I are looking for a few testers for something special we've been building! It's an app that lets you view the same content you see here on Bluesky but with extra features — some focused on researchers, some general upgrades. Reach out to us if interested!
December 29, 2025 at 6:55 PM
@eugenevinitsky.bsky.social and I are looking for a few testers for something special we've been building! It's an app that lets you view the same content you see here on Bluesky but with extra features — some focused on researchers, some general upgrades. Reach out to us if interested!
Reposted by Nikhil Garg
Researchers: Aside from moderation and safety features, what would you want in a researcher-focused social app? Obviously a feed of shared papers, but what else?
We could make our own bsky client that does this by default? Maybe?
December 24, 2025 at 6:43 PM
Researchers: Aside from moderation and safety features, what would you want in a researcher-focused social app? Obviously a feed of shared papers, but what else?
Reposted by Nikhil Garg
I keep telling this to students. If you really want to do SAE research, just find a great classic paper on topic modeling and reimplement it using SAEs. I don’t think anyone would really notice.
Interpretable Embeddings with Sparse Autoencoders: A Data Analysis Toolkit
Analyzing large-scale text corpora is a core challenge in machine learning, crucial for tasks like identifying undesirable model behaviors or biases in training data. Current methods often rely on cos...
arxiv.org
December 15, 2025 at 10:28 PM
I keep telling this to students. If you really want to do SAE research, just find a great classic paper on topic modeling and reimplement it using SAEs. I don’t think anyone would really notice.
Reposted by Nikhil Garg
And we have lots of open roles at ACLU, including roles on our Technology Team: www.aclu.org/careers/
Careers at ACLU
Join our team! We’re looking for committed, passionate people for open roles at the ACLU.
www.aclu.org
December 14, 2025 at 1:46 AM
And we have lots of open roles at ACLU, including roles on our Technology Team: www.aclu.org/careers/
Reposted by Nikhil Garg
Lots of new roles just went up at Code for America! codeforamerica.org/jobs/
Careers — Code for America
Our team is made up of empathetic people working side by side with communities and government to solve society's toughest problems
codeforamerica.org
December 14, 2025 at 1:43 AM
Lots of new roles just went up at Code for America! codeforamerica.org/jobs/
Reposted by Nikhil Garg
Most LLM evals use API calls or offline inference, testing models in a memory-less silo. Our new Patterns paper shows this misses how LLMs actually behave in real user interfaces, where personalization and interaction history shape responses: arxiv.org/abs/2509.19364
December 12, 2025 at 8:42 PM
Most LLM evals use API calls or offline inference, testing models in a memory-less silo. Our new Patterns paper shows this misses how LLMs actually behave in real user interfaces, where personalization and interaction history shape responses: arxiv.org/abs/2509.19364
We did oral exams for my class this fall too. A lot of time but worth it I think.
The strong winds of a “back to basics” movement are blowing, between this trend and the book reading complaint (and the grade inflation one, and the testing one, and the accountability…). It’s coming.
It's been a super rewarding semester of teaching this fall!
I spoke with the Washington Post about administering oral exams for the first time in my Intro Data Science class - to combat AI use. Happy to chat with folks thinking about doing something similar:
www.washingtonpost.com/education/20...
I spoke with the Washington Post about administering oral exams for the first time in my Intro Data Science class - to combat AI use. Happy to chat with folks thinking about doing something similar:
www.washingtonpost.com/education/20...
December 12, 2025 at 4:03 PM
We did oral exams for my class this fall too. A lot of time but worth it I think.
Attended CODE for the first time this year and it was great
If you missed this year's Conference on Digital Experimentation, the plenary talks are now available!
Videos from this year's plenary talks on digital experimentation are now available: www.youtube.com/@mitide/videos
December 8, 2025 at 2:42 PM
Attended CODE for the first time this year and it was great
Reposted by Nikhil Garg
ever since i read this i’ve been convinced that the uk is basically being governed by chat
www.nytimes.com/2025/12/03/m...
www.nytimes.com/2025/12/03/m...
December 7, 2025 at 10:08 AM
ever since i read this i’ve been convinced that the uk is basically being governed by chat
www.nytimes.com/2025/12/03/m...
www.nytimes.com/2025/12/03/m...