I gave a talk a month ago on support recovery in 1 bit compressed sensing, where a very curious open problem remains even after some serious attempts. Support recovery with few false positives can be done with (sparsity)^1.5 measurements. Is that optimal? youtube.com/watch?v=OndWAU…

Suppose you want to detect if there is unusual dense part in a graph (say, coauthor network). Getting the whole graph is however costly (memory, access), so you try to see as small a part of the graph as possible. By Wasim Huleihel and Soumyabrata Pal arxiv.org/abs/2110.00744

This is the kind of heart, the kind of attitude that is truly inspiring!

Edwin Lord Weeks, a famed painter, captured India's vibrant landscapes and rich culture from 1882-1888. I planned 2 hours for this thread but spent 6, mesmerized. Zoom in the find intricacies and share your thoughts on Sunday!! #MegaThread 1/29

From Bhala Kyaa Uthayega to Gold Medal Jab khelne chala tha yeh, Hasi chooti thi unki. Bhaala kya uthayega yeh, Size ki di thi dhamki. Aa raha hai Navdeep, apne khel se teri hasi mitane. Duniya ko batane, ke Toofan wahi, jo #TaanePalatDe

A student reached out asking for advice on research directions in optimization, so I wrote a long response with pointers to interesting papers. I thought it'd be worth sharing it here too: 1. Adaptive optimization. There has been a lot going on in the last year, below are some

"Writing papers with fevers, solving research problems in the hospital, reading papers during treatment". Nami Matsumoto did a q/a in the EnCORE Institute magazine. I will be very happy if more folks read this: encore.ucsd.edu/encore-magazin… p.20. Very proud of her and her work.

🥳Our workshop proposal, "Building Trust in Language Models and Applications," has been accepted by ICLR 2025! Shout out to all the organizers Micah Goldblum ramasuri narayanam Soumyabrata Pal Martin Pawelczyk (on Job Market) 𝙷𝚒𝚖𝚊 𝙻𝚊𝚔𝚔𝚊𝚛𝚊𝚓𝚞 and Shiv Kumar Saini! openreview.net/forum?id=IiTBR…

A mathematical proof finally explains why quantizing LLMs layer-by-layer actually works. This paper introduces a theoretical framework linking layer-wise quantization error to model perplexity, enabling better data-free quantization methods for LLMs. The framework proves a

The final moments of the #DingGukesh World Championship Match in 11 Languages! ♟️

(1/4) Daniel Markovits, a Yale Law School Prof, argues that meritocracy focuses on superiority rather than excellence. This is a critically important point that applies not only to society at large, but to how we measure progress in science - and AI in particular - these days.

Link: building-trust-in-llms.github.io/iclr-workshop/… Thanks to co-organizers Bang An, ramasuri narayanam, Soumyabrata Pal, Martin Pawelczyk (on Job Market), 𝙷𝚒𝚖𝚊 𝙻𝚊𝚔𝚔𝚊𝚛𝚊𝚓𝚞, and Shiv Kumar Saini

✅ Silver at the Asian Championships 🥈 ✅ Back-to-back Personal Bests 👏 ✅ Second medal in 3000m Steeplechase at the Asian Championships Parul Chaudhary began her 2025 season in stunning fashion at the Doha Diamond League, clocking 9:13.39 to better her own National Record in

🚨Pooja’s torn spikes, bandaged together — and still a golden leap! 🇮🇳🥇 On Day 1 of #AsianAthletics2025, Pooja’s spikes tore during practice. But instead of giving up, she wrapped them with bandage, held them together — and leapt her way into history! She cleared 1.89m,

INDIA HAVE QUALIFIED TO ASIAN CUP! WE ARE IN TEARS OF JOY! 🇮🇳😭🔥❤️

🗣️ Sangita Basfore: It's not just about the Asian Cup! The World Cup is our target. 💫 #WAC2026 #BlueTigresses #IndianFootball ⚽️

INDIA QUALIFY TO U20 ASIAN CUP AFTER 20 YEARS!!!!! 😭❤️🇮🇳🔥

This is brilliant 🤩👌🏼👏🏼👏🏼 You will enjoy