🚨 New Paper 🚨 It’s about time someone unraveled the mystery of date/time bugs affecting modern software systems and how to catch them! 📢 I'm excited to share our upcoming MSR'25 paper on exactly this. 📜 Paper: rohan.padhye.org/files/datetime… Thread...

What design decisions in LLM training affect the final performance of LLMs? Scaling model size and training data is important, but it's not the only thing. We performed an analysis of 90+ open-weights models to answer this question. 🧵 arxiv.org/abs/2503.03862 (1/12)

📢 New Paper! Tired 😴 of reasoning benchmarks full of math & code? In our work we consider the problem of reasoning for plot holes in stories -- inconsistencies in a storyline that break the internal logic or rules of a story’s world 🌎 W/ Melanie Sclar, and tsvetshop 1/n

SIGBOVIK 2025 is out A live thread of papers that make me lightly smile: 1/ UPPERCASE IS ALL YOU NEED

Excited to announce our #NAACL2025 Oral paper! 🎉✨ We carried out the largest systematic study so far to map the links between upstream choices, intrinsic bias, and downstream zero-shot performance across 131 CLIP Vision-language encoders, 26 datasets, and 55 architectures!

At #NAACL2025 presenting the poster for our work! 📍 Hall 3, Session B on 🗓️ Wednesday, April 30 (tomorrow!) 🕚 11:00–12:30 Stop by to discuss the subjective standards of data selection in instruction following!

Extremely happy to have our work on Teaching Transformers Causal Reasoning through Axiomatic Training accepted at ICML!!! Very proud of this work and the potential avenues it opens up on improving causal reasoning capabilities of LLMs.

New paper alert! 🚀📈 #ICML2025 ICML Conference Can graph curvature uncover hidden anomalies overlooked by traditional methods? 💡Introducing CurvGAD, a mixed-curvature graph autoencoder designed to detect curvature-based geometric anomalies. 🔗 arxiv.org/pdf/2502.08605 (1/9)

Correction: The poster session is going to be in Hall 3, Boards 10-19.

Thrilled to launch Prompt Adaptation, a state-of-the-art agentic system to automate prompt engineering 🚀

So honoured to be selected by ⁦Forbes India⁩ India & ⁦Accel⁩ amongst the 30 Indians who are leading AI And as always, this is a recognition of the all the people I have worked with, who have supported me and set me up to do my best linkedin.com/posts/forbes-i…

When it comes to text prediction, where does one LM outperform another? If you've ever worked on LM evals, you know this question is a lot more complex than it seems. In our new #acl2025 paper, we developed a method to find fine-grained differences between LMs: 🧵1/9

Excited to be in Vancouver this week for #ICML2025!🌍✈️. Presenting our paper on curvature-based graph anomaly detection: 📍 East Exhibition Hall A-B E2900 📅 Tue, 15 July | 4:30 – 7:00 PM PDT 🔗 arxiv.org/abs/2502.08605 Drop by to chat about #geometricDL and #graphML! 🚀🧩

Can we create effective watermarks for LLM training data that survive every stage in real-world LLM development lifecycle? Our #ACL2025Findings paper introduces fictitious knowledge watermarks that inject plausible yet nonexistent facts into training data for copyright

Wow! Prof Mitesh Khapra's name is in Times 100/AI in the same list as Fei-Fei Li, Sam Altman, Elon Musk, Jeff Dean and so on. I bless the day I was invited to work alongside him in AI4Bharat! Truly a visionary! time.com/collections/ti… Reply 🐐 to show respect!

Excited to share that I’ve joined Meta as a Visiting Researcher for the next year, in their Pittsburgh office, alongside my PhD at Machine Learning Dept. at Carnegie Mellon. I’ll be working on #GeometricDL and #GraphML, with Yinglong Xia & Hanqing Zeng, under the AI at Meta AI Mentorship Program.

A lot is said about LLMs’ counterfactual reasoning, but do they truly possess the cognitive skills it needs? Introducing Executable Counterfactuals, a code framework that (1) shows frontier models lack these skills (2) offers a testbed for improvement via Reinforcement Learning

🧵Come check out our Thursday morning poster at Conference on Language Modeling - 11AM - Poster #15! “Mitigating Modal Imbalance in Multimodal Reasoning” (1/n)

As good a time as any to announce I'm on the job market this year! I develop provably efficient reinforcement learning algorithms that are directly applicable to problems across both robotics and language modeling. See gokul.dev for more!

Honored to be named an Amazon AI Ph.D. Fellow (a.y. 2025–27) ✨ Grateful to my advisors at Machine Learning Dept. at Carnegie Mellon — Christos F. and Geoff G. — for their constant support. We’ll be working on geometric DL for graphs, in close collaboration with the researchers at Amazon. #AmazonAIFellowship