Excited to unveil our latest findings in the paper "Prejudice and Volatility of LLMs" (arxiv.org/abs/2402.15481)! 🦠Data Toxicity ⬆️ → 🔵Prejudice ⬆️ 🔴Volatility ⬇️ 🐋Model Size ⬆️ → 🔵Prejudice ⬆️ 🔴Volatility ⬇️ 🧑‍🏫RLHF ☑️ → 🔵Prejudice ⬇️ 🔴Volatility ⬆️

🔍 New Preprint! Why do LLMs generate hallucinations even when trained on all truths? 🤔 Check out our paper [arxiv.org/abs/2407.08039] 💡 We find that universally, data imbalance causes LLMs to over-generalize popular knowledge and produce amalgamated hallucinations. 📊

Event-based Knowledge Editing has been accepted to EMNLP2024！

👾 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

We're also excited to announce our #EMNLP2024 panel on Wed., Nov. 13! Title: The Importance of NLP in the LLM Era featuring Heng Ji Rada Mihalcea Alice Oh Sasha Rush moderated by Monojit Choudhury

Looking forward to hearing about the amazing panel！

I am at #EMNLP2024! I will present our work "Why Does New Knowledge Create Messy Ripple Effects in LLMs? " on Web 10:30am. Thanks to all the collaborators Heng Ji Zixuan Zhang Chi Han Manling Li Looking forward to have a chat! Paper Link: arxiv.org/pdf/2407.12828

Presenting our Event-based Knowledge Editing paper at EMNLP!! Take a look at: aclanthology.org/2024.emnlp-mai…

It’s so great to collaborate with such a wonderful team! Thank you all for helping me with the work!

Transferring UIUC Spring Apartments at an Extremely Low Price❗️❗️ Current price $500 per month vs. Original price $890 for one room in Yugo 3rd Lofts 2b1b 🎮🌞🎮 Reach out to know more details if you are interested. 💨💨💨

Excited to share 1/2 of my #coling2025 papers: PropaInsight: Toward Deeper Understanding of Propaganda! Huge thanks to my coauthors Jiateng Liu, May Fung and team, and special thanks to Julia Hirschberg, Heng Ji, Preslav Nakov for their support! Read here: arxiv.org/pdf/2409.18997

Here’s 2/2 of my #coling2025 papers: NoVAScore🌟 We introduce an automated metric to assess document novelty and salience. Huge thanks to my coauthor Ziwei (Sara) Gong and the team for their amazing collaboration! Read here: arxiv.org/pdf/2409.09249

📢📈 I’m on the 2025 faculty job market! I've been incredibly grateful to work with inspiring advisors, mentors & peers. 💡My research, AI4Scientists🔬, accelerates & democratizes the research lifecycle by: 1️⃣ Few-shot scientific knowledge acquisition 2️⃣ Domain-aware scientific

Proud to announce my recent work on domain specific language modeling! We present a cost-effective methodology for filtering large, general corpus datasets to a specific domain, like astronomy, and training a LLM on the result for great improvements. arxiv.org/abs/2412.14436

🙌 Happy New Year everyone! 🤖 New preprint: TinyHelen's First Curriculum: Training and Evaluating Tiny Language Models in a Simpler Language Environment 🤖 We train and evaluate tiny language models (LMs) using a novel text dataset with systematically simplified vocabularies and

Why allocate the same number of visual tokens to a blank image and a complex landscape? Introducing DyMU: a training-free algorithm that makes any ViT visual encoder dynamic-length and plug-and-play with downstream VLMs. 🚀 🔗 Project Page: mikewangwzhl.github.io/dymu/

We are extremely excited to announce mCLM, a Modular Chemical Language Model that is friendly to automatable block-based chemistry and mimics bilingual speakers by “code-switching” between functional molecular modules and natural language descriptions of the functions. 1/2

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.

Multimodal conversational agents struggle to follow complex policies, which also impose a fixed computational cost. We ask: 👉 How can we achieve stronger policy-following behavior without having to include policies in-context? 🌐: mikewangwzhl.github.io/TriMPI/ 🧵1/3