Shiyu Chang

@CodeTerminator

Associate Professor of CS @ UCSB.

Santa Barbara, CA

code-terminator.github.io

Joined October 2016

Shiyu Chang
@CodeTerminator
Jul 15, 2025
Sad to miss #ICML2025 this year, but thrilled that my student @hou_bairu will present his exciting work on dynamically pruning LLMs into efficient, task-specific models—done in collaboration with our amazing collaborators from Apple! 🍎✨
Bairu Hou
@hou_bairu
Jul 15, 2025
Just describe your task (and optionally the input) — our method then dynamically prune the LLM into a smaller model that’s tailor-made for the task/input and gets it ready for inference in just 0.1 seconds. We call it "instruction-following" model pruning. Check out our
Shiyu Chang
@CodeTerminator
Jun 14, 2025
Our new knowledge tracing method accurately tracks young student understanding — even with just a few student responses. Led by my amazing students @xiny_gao @__wuqiuche__, with fantastic collaborators! 🌱🧠
Xinyi Gao
@xiny_gao
Jun 14, 2025
1/3 How can we reliably trace a student’s understanding with just a few exercise responses? In classrooms, it’s important for teachers to track students’ understanding and predict their future performance on exercises. However, existing knowledge tracing (KT) algorithms often
GIF
Shiyu Chang
@CodeTerminator
May 1, 2025
Sadly missing #NAACL this time, but my student @hou_bairu will present our joint work with @nlp_mit and @MITIBMLab on hallucination detection using a novel probabilistic propagation approach — one of my favorite recent projects from our team.
Bairu Hou
@hou_bairu
May 1, 2025
Excited to share our #NAACL 2025 work on detecting LLM hallucinations! We, @YangZha26484161 , @jacobandreas , @CodeTerminator , propose a novel method: Given an LLM-generated text, first generate logical-related statements and organize them in a tree structure. Then we probe the
Shiyu Chang
@CodeTerminator
Apr 23, 2025
Sad to miss #ICLR2025 this year, but thrilled to see our work led by my student @liu_yujian presented there! Stop by our poster and chat with my amazing collaborators about Prereq-Tune, our new method to improve LLM factuality. 🚀📍#273 | Today (Apr 24), 3–5:30 PM
Yujian Liu
@liu_yujian
Apr 23, 2025
How can we train LLMs to say only what they truly know? Our #ICLR2025 paper introduces Prereq-Tune, a novel fine-tuning strategy that: 1️⃣ Disentangles the learning of knowledge and skills via two-stage training. 2️⃣ Uses fictitious synthetic data to precisely control what the
Shiyu Chang
@CodeTerminator
Apr 22, 2025
Elegant algorithm to embed watermarks against piggyback spoofing attacks! 🚀 Led by my amazing student @LiAn1551016 and our great collaborators.
Li An
@LiAn1551016
Apr 22, 2025
You’re an LLM provider. Someone inserts toxic content into your watermarked output — and your watermark still says it’s yours. 😨 That’s a spoofing attack. How do you defend trust? We propose a novel watermarking method that fights back: 🛡️ 🟩 Robustness: watermark persists