Publications - Jingxuan Wu

← Back to Home

Publications

(* indicates equal contribution)

2026
Preprint	Judged Useless, Queried Anyway: Tool-Using Agents Rarely Turn Their Own Evidence Judgments into Stopping Decisions Chubin Zhang, Zhenglin Wan, Xingrui Yu, Jingxuan Wu, Yaxin Zhou, Ivor Tsang, Bo An Shows that tool-using agents often recognize useless evidence but fail to turn that judgment into stopping decisions. Paper
Preprint	MemFold: Learning Compact Soft Memory for Long-Context Personalization via On-Policy Optimization Jingxuan Wu, Yuzhe Yang, Yiqiao Huang, Chengzhi Liu, Qingni Wang, Chengxuan Qian, Shutong Wu, Jiawei Zhang, Xin Eric Wang Learns compact soft memories for long-context personalization through on-policy optimization. Paper / Code / Project
In Preparation	SafeScreen: Bias-Safe AI Screening for Adaptive Visual Conjoint Jingxuan Wu, Guanting Chen, Wenjia Ba A bias-safe AI screening framework for adaptive visual conjoint analysis.
Preprint	Don't Blindly Trust It: How Unreliable Feedback Breaks Tool-Using LLM Agents Chubin Zhang, Zhenglin Wan, Xingrui Yu, Pengfei Zhou, Wangbo Zhao, Jingxuan Wu, Yaxin Zhou, Ivor Tsang Studies how unreliable environmental feedback can derail tool-using LLM agents. Paper
NeurIPS 2026	Calibration Is Not Control: Why LLM-Agent Oversight Needs Intervention Chubin Zhang, Zhenglin Wan, Xingrui Yu, Jingxuan Wu, Qi Wen, Pengfei Zhou, Wangbo Zhao, Ivor Tsang Shows why calibrated uncertainty alone is insufficient for effective LLM-agent oversight. Paper
NeurIPS 2026	Denoising Time Matters: Diverse Generation in Diffusion Language Models Jingxuan Wu, Zhenglin Wan, Xingrui Yu, Yuzhe Yang, Yiqiao Huang, Ivor Tsang, Yang You Also accepted at the ICML 2026 Workshop on SPIGM. Improves diversity in diffusion language models through time-annealed perturbations while preserving generation quality. Project / Paper / Code
Technical Report	CaveAgent: Transforming LLMs into Stateful Runtime Operators Maohao Ran, Zhenglin Wan, Cooper Lin, Yanting Zhang, Hongyu Xin, Hongwei Fan, Yibo Xu, Beier Luo, Yaxin Zhou, Wangbo Zhao, Lijie Yang, Lang Feng, Fuchao Yang, Jingxuan Wu, (etc..) Bo An, Yike Guo, Jun Song Turns LLMs into stateful runtime operators for reliable long-horizon task execution. Paper / Code / Website
2025
ICML 2026	Adversarial Dual On-Policy Distillation from Expressive Teacher Zhenglin Wan, Jingxuan Wu, Xingrui Yu, Chubin Zhang, Mingcong Lei, Bo An, Ivor Tsang Distills expressive teachers into efficient policies through adversarial dual on-policy learning. Paper / Code
ICML 2026	Letting Trajectories Spread: Quality-Preserving Control for Diverse Flow Matching Jingxuan Wu, Zhenglin Wan, Xingrui Yu, Yuzhe Yang, Bo An, Ivor Tsang Improves sample diversity in flow matching while preserving generation quality. Paper / Code / Project
2024
In Preparation	LLM-Powered Predictive Decision-Making for Sustainable Data Center Operations Hanzhao Wang, Jingxuan Wu, Yu Pan, Yumeng Li, Guanting Chen Combines LLM forecasting and optimization for sustainable data-center decision-making. Paper
TR-E	Flight Delay Propagation Modeling: Data, Methods, and Future Opportunities Chi Li, Jianfeng Mao, Lingyi Li, Jingxuan Wu, Lianmin Zhang, Jianyu Zhu, Zibin Pan Reviews data and modeling approaches for understanding flight-delay propagation. DOI
In Preparation	Integrative Mean-Field Epidemic Model and Adaptive Graph Learning for Network-wide Delay Propagation Dynamics Prediction Chi Li, Mingcong Lei, Jingxuan Wu, Yuzhe Yang, Zibin Pan, Xiongwen Qian, Jianfeng Mao Combines mean-field epidemic dynamics with adaptive graph learning for network-wide delay prediction. Paper