Life update: Excited to announce that I’ll be joining UCSB CS (UCSB Computer Science Department) as an Assistant Professor in Fall 2025—where my research dream began! If you’re interested in PhD research (3+ openings) on multimodal, GenAI, or agents (embodied/digital), apply to UCSB CS by Dec 15! 1/n

World models must enable interaction, allowing agents to observe state changes after interventions and actions. Otherwise, they’re no different from standard predictive models 🧐. Currently, technologies like pure video generation and some 2D-to-3D transformation still lack

Excited to announce our speculative knowledge distillation (arxiv.org/abs/2410.11325) is accepted to ICLR 2026 ICLR2025! In this work, we addressed both limitations of supervised and on-policy KD, leading to a new SOTA KD performance! Many thanks to Google cloud AI research team!

🚀 Excited to share that our work has been accepted to #NAACL2025! We show that LLM watermarks can be removed in a black-box setting 🛠️ For more details: arxiv.org/abs/2411.01222

Amazing

Nice work!

Gave an oral presentation at #NAACL2025 today! Graceful for the chance to share our work on text watermark removal, and happy to connect with old friends and meet new ones here!

Excited to share our paper at #ICML2025! We've developed MELON🍉, a robust defense method against indirect prompt injection attacks on LLM agents that achieves near 0 ASR! Hope you enjoy🍉! Grateful to my incredible collaborators! William Wang Wenbo Guo Jindong Wang Xianjun Yang

If you're trying to decide which CVPR tutorial to attend, Check out Kaijie Zhu, Sophia Xiao Pu, @nieyuzhou, Hao Chen, and my tutorial on Evaluating Multimodal Models Unfortunately, I won't be able to attend, but my lovingly crafted slides are in the mix :)

CVPR in Nashville: vision models by day, audio models by night (aka live music on every corner 🎸). Multimodal IRL 😎 #CVPR2025

Huge congratulations to the new doctors! 🎉

👀 How to find more difficult/novel/salient evaluation data? ✨ Let the data generators find it for you! Introducing Data Swarms, multiple data generator LMs collaboratively search in the weight space to optimize quantitative desiderata of evaluation.

We did it! 🎉 12 papers from UCSB NLP accepted at #EMNLP2025 (7 Main + 5 Findings) Proud of everyone’s hard work—poster below 👇