Lyuxin David Zhang @ COLM

Undergrad CS & Math @Penn, ML & NLP @cogcomp @.SGTM @upennnlp #YNWA

[Daniel Fried @dan_fried]

Our project builds on past work with many collaborators, especially and most recently @uilydna , @saujasv, @aydan_huang265, and @Hongyicai2002

[Anton Xue @AntonXue]

Grateful to have worked on this with such incredible students @rah4927 and @ByAlanLi!

AI today is solving open problems left and right, but human steering still matters for efficiently identifying what's interesting + promising.

[David Wan @meetdavidwan]

🥳 Excited to share that MuRGAt is accepted to #ICML2026!

Even strong MLLMs hallucinate citations to multimodal sources (video, audio, charts). Our new Fact-Level Multimodal Attribution benchmark tackles this by:
🕐 Requiring fine-grained temporal & per-modality citations (vs. just source-level)
🔍 Distinguishing verifiable claims from reasoning steps to evaluate multi-step responses

We also introduce MuRGAt-SCORE, a reference-free, decomposed metric aligned with human judgment, and show that Programmatic Grounding substantially boosts attribution!

👇

[Vivek Gupta @keviv9]

Our ACL 2026 Finding paper, show that decomposed prompting can help: https://t.co/4bgPz0O6jR

Decomposed Prompting Does Not Fix Knowledge Gaps, But Helps Models Say "I Don't Know"

with @DanRothNLP, @TomerWolfson

[Tomer Wolfson @TomerWolfson]

Many factual QA benchmarks have become saturated, yet factuality still poses a very real issue!

✨We present MoNaCo, an Ai2 benchmark of human-written time-consuming questions that, on average, require 43.3 documents per question!✨

📣Blogpost: https://t.co/GQD83gdHgg

🧵(1/5)

[Lyuxin David Zhang @ COLM @davidzhangsan]

@gyrojeff_ 酷酷酷

[Lyuxin David Zhang @ COLM @davidzhangsan]

@Hongyicai2002 Very exciting results!

[Hongyi James Cai @Hongyicai2002]

🚀Excited to share our new paper: How Much Backtracking is Enough?

🤔 How many backtracks should your LLM learn to reason better?
Turns out: the harder the task, the more backtracking you need!

[Lyuxin David Zhang @ COLM @davidzhangsan]

@gyrojeff_ 浦东都有

[Anton Xue @AntonXue]

Excited to present our paper on a logic-based perspective of LLM jailbreaks with @Avishreekh at @ICLR_conf this Saturday, April 26!

Poster #268 in Hall 3+2B at 15:00 Singapore time
📄 arXiv: https://t.co/2wBtqvIIwD
🔗 Blog: https://t.co/f6OHxORDgb

\begin{thread}

[Lyuxin David Zhang @ COLM @davidzhangsan]

@gyrojeff_ 6

[Lyuxin David Zhang @ COLM @davidzhangsan]

@gyrojeff_ 巡回演出

[Lyuxin David Zhang @ COLM @davidzhangsan]

@gyrojeff_ 今天天气超级好啊

[Lyuxin David Zhang @ COLM @davidzhangsan]

@gyrojeff_

[Lyuxin David Zhang @ COLM @davidzhangsan]

@gyrojeff_ vienna waits for u

[Lyuxin David Zhang @ COLM @davidzhangsan]

@gyrojeff_ 老乱

[Bhuwan Dhingra @bhuwandhingra]

🧵When should LLMs trust external contexts in RAG?

New paper from @YukunHuang9 and @sanxing_chen enhances LLMs’ *situated faithfulness* to external contexts -- even when they are wrong!👇

[Dowon 'Zio' Kim @ZioKim22]

Committed! @PrincetonMBB 🟠⚫️

[Lyuxin David Zhang @ COLM @davidzhangsan]

@ZioKim22 @Nextplaybasket1 @TPSBasketball @gritbasketball1 top scholar, top baller

[Lyuxin David Zhang @ COLM @davidzhangsan]

@ZioKim22 @TPSBasketball W team, W players