Publications

Equal contribution indicated by *

2026

  1. arXiv
    Practical Online KV Cache Compaction for LLM Agents: An Empirical Study
    Yujian Liu*, Jiabao Ji*, Li An*, Rohit Jain, Gungor Polatkan, Siyu Zhu, and Shiyu Chang
    2026
    PDF
  2. EMNLP-Findings
    Speculate While You Reason: Teaching Agents to Predict Their Next Tool Call via Joint Agent-Speculator RL
    Jiabao Ji*, Yujian Liu*, Li An*, Rohit Jain, Gungor Polatkan, Siyu Zhu, and Shiyu Chang
    2026
    PDF
  3. COLM
    How Well Do Agentic Skills Work in the Wild: Benchmarking LLM Skill Usage in Realistic Settings
    Yujian Liu*, Jiabao Ji*, Li An, Tommi Jaakkola, Yang Zhang, and Shiyu Chang
    2026
    PDF Code
  4. TMLR
    Learning from Online Videos at Inference Time for Computer-Use Agents
    Yujian Liu, Ze Wang, Hao Chen, Ximeng Sun, Xiaodong Yu, Jialian Wu, Jiang Liu, Emad Barsoum, Zicheng Liu, and Shiyu Chang
    2026
    PDF Code
  5. TMLR
    ThinkPrune: Pruning Long Chain-of-Thought of LLMs via Reinforcement Learning
    Bairu Hou, Yang Zhang, Jiabao Ji, Yujian Liu, Kaizhi Qian, Jacob Andreas, and Shiyu Chang
    2026
    PDF Code
  6. arXiv
    WorldAuditBench: Interactive 3D World Auditing with Multimodal Agents
    Ziyan Jiang*, Jingbo Yang*, Jiabao Ji*, Yujian Liu, Qiucheng Wu, Tommi Jaakkola, Yang Zhang, and Shiyu Chang
    2026
    PDF Code
  7. arXiv
    VISUALSKILL: Multimodal Skills for Computer-Use Agents
    Ziyan Jiang*, Li An*, Yujian Liu, Jiabao Ji, Qiucheng Wu, Jacob Andreas, Yang Zhang, and Shiyu Chang
    2026
    PDF Code

2025

  1. DL4C@NeurIPS
    HarnessLLM: Automatic Testing Harness Generation via Reinforcement Learning
    Yujian Liu, Jiabao Ji, Yang Zhang, Wenbo Guo, Tommi Jaakkola, and Shiyu Chang
    2025
    PDF Code
  2. ICLR
    Fictitious Synthetic Data Can Improve LLM Factuality via Prerequisite Learning
    Yujian Liu, Shiyu Chang, Tommi Jaakkola, and Yang Zhang
    2025
    PDF Code
  3. EMNLP-Findings
    Augment before You Try: Knowledge-Enhanced Table Question Answering via Table Expansion
    Yujian Liu, Jiabao Ji, Tong Yu, Ryan Rossi, Sungchul Kim, Handong Zhao, Ritwik Sinha, Yang Zhang, and Shiyu Chang
    2025
    PDF Code
  4. arXiv
    A Reinforcement Learning Framework for Robust and Secure LLM Watermarking
    Li An, Yujian Liu, Yepeng Liu, Yuheng Bu, Yang Zhang, and Shiyu Chang
    2025
    PDF Code
  5. COLM
    Defending LLM Watermarking Against Spoofing Attacks with Contrastive Representation Learning
    Li An, Yujian Liu, Yepeng Liu, Yang Zhang, Yuheng Bu, and Shiyu Chang
    2025
    PDF Code

2024

  1. EMNLP
    Revisiting Who’s Harry Potter: Towards Targeted Unlearning from a Causal Intervention Perspective
    Yujian Liu, Yang Zhang, Tommi Jaakkola, and Shiyu Chang
    2024
    Oral
    PDF Code
  2. NeurIPS
    Reversing the Forget-Retain Objectives: An Efficient LLM Unlearning Framework from Logit Difference
    Jiabao Ji, Yujian Liu, Yang Zhang, Gaowen Liu, Ramana Rao Kompella, Sijia Liu, and Shiyu Chang
    2024
    PDF Code
  3. ICML
    Decomposing Uncertainty for Large Language Models through Input Clarification Ensembling
    Bairu Hou, Yujian Liu, Kaizhi Qian, Jacob Andreas, Shiyu Chang, and Yang Zhang
    2024
    Oral (top 1.5% of submissions)
    PDF Code
  4. CVPR
    Correcting Diffusion Generation through Resampling
    Yujian Liu, Yang Zhang, Tommi Jaakkola, and Shiyu Chang
    2024
    Highlight (top 2.8% of submissions)
    PDF Code

2023

  1. ICCV
    Harnessing the Spatial-Temporal Attention of Diffusion Models for High-Fidelity Text-to-Image Synthesis
    Qiucheng Wu*, Yujian Liu*, Handong Zhao, Trung Bui, Zhe Lin, Yang Zhang, and Shiyu Chang
    2023
    PDF Code
  2. CVPR
    Uncovering the Disentanglement Capability in Text-to-Image Diffusion Models
    Qiucheng Wu, Yujian Liu, Handong Zhao, Ajinkya Kale, Trung Bui, Tong Yu, Zhe Lin, Yang Zhang, and Shiyu Chang
    2023
    PDF Code Website
  3. EMNLP
    All Things Considered: Detecting Partisan Events from News Media with Cross-Article Comparison
    Yujian Liu, Xinliang Frederick Zhang, Kaijian Zou, Ruihong Huang, Nicholas Beauchamp, and Lu Wang
    2023
    PDF
  4. ICWSM
    Improving Mental Health Classifier Generalization with Pre-Diagnosis Data
    Yujian Liu*, Laura Biester*, and Rada Mihalcea
    2023

2022

  1. NAACL Findings
    POLITICS: Pretraining with Same-story Article Comparison for Ideology Prediction and Stance Detection
    Yujian Liu*, Xinliang Frederick Zhang*, David Wegsman, Nicholas Beauchamp, and Lu Wang
    2022
    PDF Code Demo

2020

  1. arXiv
    Audrey: A Personalized Open-Domain Conversational Bot
    Chung Hoon Hong, Yuan Liang, Sagnik Sinha Roy, Arushi Jain, Vihang Agarwal, Ryan Draves, Zhizhuo Zhou, William Chen, Yujian Liu, Martha Miracky, and 3 more authors
    2020
    PDF