💡In the era of pretrained LLM backbones, what are fundamental mechanisms behind effectively & efficiently controlling LM generation style? 🚀 Introducing “Word Embeddings Are Steers for Language Models”, which we'll present at #ACL2024! 📜 Preprint: arxiv.org/pdf/2305.12798

Super honored and humbled to have been a recipient of the #NAACL2024 Outstanding Paper award for our R-Tuning paper. A big shoutout of congratz & thanks to the lead authors Hanning+Shizhe+Yong, and our entire team at @UIUC and @HKUST! 🤗

Want to hear more about "The Role of Semantics in the Age of LLMs"? 📈 For those who’re attending NAACL HLT 2025 and still around the area, check out the SEM 2024 Panel Talk this afternoon where Greg Durrett, Aline Villavicencio, Violet Peng, and I will be featured! 🌟

We have won yet another Outstanding Paper Award at ACL2024 for the LM-Steer work, again led by the most amazing Chi Han Chi Han ! His LM-Infinite work has just won NAACL2024 Outstanding Paper Award! Chi will be on academic job market next year! blender.cs.illinois.edu/paper/lmsteer2…

👾 Introducing AgentOccam: Automating Web Tasks with LLMs! 🌐 AgentOccam showcases the impressive power of Large Language Models (LLMs) on web tasks, without any in-context examples, new agent roles, online feedback, or search strategies. 🏄🏄🏄 🧙 Link: arxiv.org/abs/2410.13825

Imagine AI assistants on your smart glasses or laptop proactively bridging your info gaps! 🗺️ Entering a building? Get instant floor plans for seamless navigation. 🧑‍💻 In lectures? Receive concise explanations to stay on track. Our new preprint introduces Just-In-Time

🧠Let’s teach LLMs to learn smarter, not harder💥[arxiv.org/pdf/2506.06972] 🤖How can LLMs verify complex scientific information efficiently? 🚀We propose modular, reusable atomic reasoning skills that reduce LLMs’ cognitive load to verify scientific claims with little data.

🧠 How can AI evolve from statically 𝘵𝘩𝘪𝘯𝘬𝘪𝘯𝘨 𝘢𝘣𝘰𝘶𝘵 𝘪𝘮𝘢𝘨𝘦𝘴 → dynamically 𝘵𝘩𝘪𝘯𝘬𝘪𝘯𝘨 𝘸𝘪𝘵𝘩 𝘪𝘮𝘢𝘨𝘦𝘴 as cognitive workspaces, similar to the human mental sketchpad? 🔍 What’s the 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗿𝗼𝗮𝗱𝗺𝗮𝗽 from tool-use → programmatic