📣We are hiring! My group at UW–Madison Computer Sciences has open positions for Postdoc and PhD starting in Fall 2025. If you're passionate about advancing responsible AI and understanding LLMs/MLLMs from both theoretical and empirical perspectives, we’d love to hear from you. Curious about

🌟 I will attend the Rising Stars in Data Science workshop Halıcıoğlu Data Science Institute Stanford Data Science Data Science Institute from Nov 14-15! Honored to be among emerging DS researchers. Looking forward to connecting with fellow participants, faculty, and industry leaders. If you'll be there, let’s meet up!

I will attend NeurIPS Conference from 12/10-12/14! 🌟I am on the academic job market, looking for assistant professor position in CS,ECE,DS, etc. My research is in reliable ML and foundation models (Statements at d12306.github.io)! Would love to chat on this during conference!

Finally happening - I am flying out to Vancouver for NeurIPS Conference this week! We are presenting several new works on hallucination detection, LLM safety, robust adaption of VLM, and more. Please stop by and check out these works led by my students Sean Xuefeng Du Hyeong-Kyu Froilan Choi

Our paper "Your Weak LLM is Secretly a Strong Teacher for Alignment" has been accepted to #ICLR2025. openreview.net/forum?id=sGqd1… Congratulations to Leitian Tao!

How should we assign rewards to intermediate steps in reasoning? DeepSeek-R1 paper highlights it as an open challenge. Here’s everything you need to know about Process Reward Models—the progress, our latest breakthrough Process Q-value Model (PQM), and how it advances OpenAI's

‼️How well do steering vectors work? When do they fail and why? ✅We perform an evaluation of steering methods and provide theoretical results explaining the results. Paper: arxiv.org/abs/2502.02716 (w/ Sharon Y. Li) [1/n]

The inflation of LLM performance metrics is a real concern. How can we quantify the degree of dataset contamination (i.e., overlap between the pre-training and benchmark evaluation data)? Dataset contamination means models might be recalling answers from pre-training data

I received many warmhearted messages from friends and colleagues these days--thank you, everyone. Receiving this award from Sloan Foundation still feels unreal. I almost didn't apply in the midst of an extremely difficult 2024 dealing with personal challenges. But I want to

🚀 New Paper Alert! 🚀 📄 Can Your Uncertainty Scores Detect Hallucinated Entity? We explore entity-level hallucination detection and benchmark 5 uncertainty-based detection methods. Paper: arxiv.org/abs/2502.11948 (w/ Max Kamachee, Seongheon Park, and Sharon Y. Li) [1/N]

Does anyone want to dig deeper into the robustness of Multimodal LLMs (MLLMs) beyond empirical observations Happy to serve this exactly through our new #ICML2025 paper "Understanding Multimodal LLMs Under Distribution Shifts: An Information-Theoretic Approach"!

MetaMind: Equip LLM with a "Social Brain" via Metacognitive Multi-agent Framework “What is meant often goes far beyond what is said, and that is what makes conversation possible.” ——H. P. Grice Huggingface: huggingface.co/papers/2505.18…

Excited to join the College of Computing and Data Science at Nanyang Technological University, Singapore (NTU Singapore) as an Assistant Professor this fall! 🙌 Grateful to my advisor Sharon Y. Li and all who supported me along the way. Looking forward to the new chapter! 😄 🇸🇬

🌍 GeoArena is live! Evaluate how well large vision-language models (LVLMs) understand the world through image geolocalization. Help us compare models via human preference — your feedback matters! 🔗 Try it now: huggingface.co/spaces/garena2… #GeoArena #Geolocation #LVLM #AI

🚨 #ICML2025 is just around the corner! I will be presenting my work on Kernel Divergence Score! 📍 East Exhibition Hall A-B #E-3012 🕚 Wed 16, 11:00 — 13:30 📄 openreview.net/pdf?id=wVDR2qm… Huge thanks to my fantastic collaborators — @khanovmax, Hongxin Wei, and Sharon Y. Li