수수

🥳Excited to share that I’ll be joining @unccs as postdoc this fall. Looking forward to work with @mohitban47 & amazing students at @unc_ai_group. I'll continue working on retrieval, aligning knowledge modules with LLM's parametric knowledge, and expanding to various modalities.

Defining “what is good” is central to both AI and Science. Despite massive efforts towards building the Virtual Cell, our recent @NatureBiotech article reflects on whether current gold standards (i.e., evaluation metrics) still hold up. nature.com/articles/s4158…

Big news! 🎉 I’m joining UNC-Chapel Hill as an Assistant Professor in Computer Science starting next year! Before that, I’ll be spending time @OpenAI working on LLM privacy. @unccs @uncnlp

I’m excited to unveil PRESAGE 🔮 for in-silico prediction of unseen single-gene perturbations on expression profiles. Developed during my Postdoc with Aviv Regev and in collaboration with @jchuetter and Jacob Levine in BRAID @ Genentech 🧬 Explore the full manuscript at…

Excited to share Video-Skill-CoT🎬🛠️– a new framework for domain-adaptive video reasoning with skill-aware Chain-of-Thought (CoT) supervision! ⚡️Key Highlights: ➡️ Automatically extracts domain-specific reasoning skills from questions and organizes them into a unified taxonomy,…

In multimodal learning, which modality interactions matter the most? 👉 Check out our new ICML 2025 paper, "I²MoE: Interpretable Interaction-aware Mixture-of-Experts", a new MoE framework that makes multimodal fusion more interpretable and effective through interaction modeling…

Thrilled to share that I’ll be joining the College of Computing and Data Science at Nanyang Technological University (NTU) (@NTUsg) as an Assistant Professor, starting in August 2025 🇸🇬🥳 I’ll continue my research on building trustworthy and continually adaptable multimodal AI,…

Sharing some personal updates 🥳: - I've completed my PhD at @unccs! 🎓 - Starting Fall 2026, I'll be joining the Computer Science dept. at Johns Hopkins University (@JHUCompSci) as an Assistant Professor 💙 - Currently exploring options + finalizing the plan for my gap year (Aug…

Starting this Fall, I’ll be joining the Computer Science Department at Emory University as an assistant professor. My group will focus on 3D computer vision and graphics, working on human digitization, 3D/4D reconstruction, manipulation, generative AI, AR/VR, and telepresence.

🔥 BIG CONGRATS to Elias + @UTAustin! Really proud of you -- it has been a complete pleasure to work with Elias and see him grow into a strong PI on *all* axes 🤗 Make sure to apply for your PhD with him -- he is an amazing advisor and person! 💙

I will be presenting ✨Reverse Thinking Makes LLMs Stronger Reasoners✨at #NAACL2025 ! We show that LLM can also benefit from reverse thinking -- a technique we often use to reason from a problem to a solution: - Improvements across 12 datasets - Outperforms SFT with 10x more…

✈️ Excited to share that I'll be presenting three (co-)first-authored papers at #ICLR2025 this Thursday and Friday! Huge thanks to all the amazing collaborators, friends, and colleagues who made this possible. 💪 Looking forward to seeing everyone at the poster sessions and…

🚨Announcing TaCQ 🚨 a new mixed-precision quantization method that identifies critical weights to preserve. We integrate key ideas from circuit discovery, model editing, and input attribution to improve low-bit quant., w/ 96% 16-bit acc. at 3.1 avg bits (~6x compression)…

🌟Meet SpatialAgent: an AI agent for spatial biology (a key tech to understand cancer and develop therapies) It rocks from experiment design to data analysis to hypothesis generation. It supports multi-organs, spans species, and can work alone or team up with human scientists🧑‍🔬

🚨 Introducing VEGGIE 🥦—a unified, end-to-end, and versatile instructional video generative model. Current video editing methods struggle with: 1. Understanding direct user instructions 2. Handling diverse editing skills in one model 3. balancing multiple training…

🎉 Excited to announce that VideoTree has been accepted to #CVPR2025 ! Check out our camera-ready for new results on: 1️⃣ Extra-long (~44min) video understanding (Video-MME, MLVU) 2️⃣ Efficiency-Effectiveness analysis: VideoTree has better performance at a lower cost 3️⃣ VideoTree…

SYMBOLIC-MOE adaptively selects expert LLMs per instance using skill-based routing. An aggregator then synthesizes outputs. 📌 Skill-based routing enables fine-grained expert selection, adapting to instance-level skill demands within tasks. 📌 Symbolic Mixture-of-Experts…

🚨Glad to share our latest work, Symbolic-MoE -- a symbolic, text-based Mixture-of-Experts framework -- using three 7-8B open-source LLMs to outperform much larger models, including GPT4o-mini and Llama 3.3 70B. Key innovations include: - Adaptiveness: Instead of using a fixed…

🚨 Excited to announce Symbolic-MoE, for efficiently recruiting+combining small (7-8B) LLMs based on their strengths/the skills needed for a query. With 3 LLMs running on 1 GPU, we beat GPT4o-mini + Llama3.3 70B, and beat multi-agent debate w/out expensive discussion. Key…

🚨 We introduce ✨ Symbolic-MoE ✨ which uses skill-based instance-level recruiting to dynamically combine LLMs, allowing three 7-8B LLMs to beat GPT4o-mini and Llama3.3 70B across challenging + diverse reasoning tasks (MMLU-Pro, AIME, GPQA, MedMCQA) while running on 1 GPU! Key…