Grateful for the huge turnout at the 1st workshop on "Time Series in the Age of Large Models" co-located with NeurIPS! Thank you to all invited speakers and participants for making this a huge success 🫶 I hope this brings forth more collaborations and better science🌐

🚀 Excited to share our latest research with you all: FM2DS: Few-Shot Multimodal Multihop Data Synthesis with Knowledge Distillation for Question Answering🌟 It addresses the challenge of multimodal question answering that require reasoning across multiple modalities and

🔥 Optimize your LLM much faster with less compute with our BlockLLM in just few lines of code (published at NeurIPS OPT 2024). 🤖 Code: lnkd.in/gBud8S27 ♠️ Thanks to my collaborators Amrutha Varshini Ramesh, Vignesh Ganapathiraman, and Mark Schmidt.

o1-preview is far superior to doctors on reasoning tasks and it's not even close, according to OpenAI's latest paper. AI does ~80% vs ~30% on the 143 hard NEJM CPC diagnoses. It's dangerous now to trust your doctor and NOT consult an AI model. Here are some actual tasks: 1/5

I’ve learned more in 1 month with ChatGPT-4o than years of courses. Here are 8 prompts that can save you thousands of dollars:

How well do you know Christmas? vancouver-ai-academy.github.io/src/xmas.html

I'm 37. When I was younger, I wasted years trapped in unresolved trauma. Then I found somatic therapy—a body-centered treatment that releases trapped emotions. Here's everything you need to know about it:🧵

It makes me ecstatic to share that InsightBench got accepted to ICLR 2026!! A huge thanks and congratulations to all the collaborators at ServiceNow Research! 🥳🎉 #ICLR2025

incredible work by my friend and colleague Perouz Taslakian

MTEB has been a great driver for text embeddings, but it's time for MMTEB, a high-quality benchmark for next-generation models! 🚀🚀🚀 1. Instruction-following embeddings 📚 2. Multilingual but still great for English 🌐 3. Highly selective tasks with less/no redundancy 🎯 4.

I’m excited to announce that 💫StarVector has been accepted at CVPR 2025! Over a year in the making, StarVector opens a new paradigm for Scalable Vector Graphics (SVG) generation by harnessing multimodal LLMs to generate SVG code that aesthetically mirrors input images and text.

super cool work by our team! A foundational SVG generation model - accepted at CVPR 2025

🚀 Internship Alert! 🇨🇦 Looking for a student to work on UI Agents & visual grounding! 🎯 🔹Develop vision-language models for UI navigation 🔹Improve visual grounding and UI understanding. Apply here: forms.gle/GWzQPhdapEVvBu… #Internship #UIAgents #ML ServiceNow Research

10 Years on and now recognized by ICLR 2026 for standing up to the test-of-time. Please join us in congratulating 🇺🇦 Dzmitry Bahdanau, Kyunghyun Cho & Yoshua Bengio for their seminal work titled “Neural Machine Translation by Jointly Learning to Align and Translate”. arxiv.org/abs/1409.0473

🚨China just officially cooked PowerPoint 🤯 PageOn AI 2.0 just launched and people are going crazy over it. This AI agent tool works like Cursor but for presentations. Here’re 4 insane features you need to know:👇🧵

A surprising & little-known results in classical statistics: Mean (μ) and median (m) are within one std deviation: |μ−m| ≤ σ For unimodal densities, bound is even tighter |μ−m| ≤ 0.7746 σ This beautiful results first appeared in a 1932 paper by Hotelling & Solomons 1/3

🚀 🚀 If you're around in Vancouver on July 13 for ICML 2025, I’m hosting a Meet & Greet Beach BBQ to kick off the conference week. 🏖️ RSVP: lu.ma/ieaubqtc Food and drinks will be provided. 📅 When: Sunday, July 13, 5:30–9:00 PM 📍 Where: English Bay Beach

New fastest shortest-path algorithm in 41 years! Tsinghua researchers broke Dijkstra’s 1984 “sorting barrier,” achieving O(m log^(2/3) n) time. This means faster route planning, less traffic, cheaper deliveries, and more efficient networks - and a CS curriculum revamp =)

getting a phd is mostly about developing a system for maximizing the time you spend in flow state, over several years

This might be the most underrated AI skill of 2025: JSON prompting. It turns your LLMs like ChatGPT, Gemini, and Claude into consistent, structured agents no hallucinations, no mess. Here’s how it works (with copy-paste templates):