Zifan Peng

Research

Problems I keep returning to

My research focuses on Trustworthy AI, Human-Centered AI, and DeFi & Blockchain, with a recent emphasis on safety and privacy for embodied AI systems.

01 9 papers Trustworthy AI

Jailbreak robustness, AI-powered search safety, privacy risk, covert content, and reliable evaluation for frontier AI systems.

 03 4 papers DeFi & Blockchain

Secure and efficient decentralized protocols, smart contract analysis, gas optimization, payment channels, and cross-chain financial primitives.

* denotes equal contribution (listed alphabetically). † denotes corresponding author. A full list is also available on Google Scholar.

Selected Papers

2026

What Your Posts Reveal: A Benchmark and Agentic Framework for User-Level Privacy Leakage on Social Media

Zifan Peng*, Yini Huang*, Aiwen Lu*, Qiming Ye, Peixian Zhang, Jingyi Zheng, Yule Liu, Xuechao Wang, Xinlei He†, Jiaheng Wei†

arXiv arXiv preprint arXiv:2606.06784
arXiv

JALMBench: Benchmarking Jailbreak Vulnerabilities in Audio Language Models

Zifan Peng, Yule Liu, Zhen Sun, Mingchen Li, Zeren Luo, Jingyi Zheng, Wenhan Dong†, Xinlei He†, Xuechao Wang, Yingjie Xue, Shengmin Xu, Xinyi Huang

ICLR International Conference on Learning Representations
arXiv

TxSum: User-Centered Ethereum Transaction Understanding with Micro-Level Semantic Grounding

Zifan Peng, Jingyi Zheng, Yule Liu, Huaiyu Jia, Qiming Ye, Jingyu Liu, Xufeng Yang, Mingchen Li, Qingyuan Gong, Xuechao Wang, Xinlei He†

arXiv arXiv preprint arXiv:2512.06933
arXiv
2025

Unsafe LLM-Based Search: Quantitative Analysis and Mitigation of Safety Risks in AI Web Search

Zeren Luo*, Zifan Peng*, Yule Liu, Zhen Sun, Mingchen Li, Jingyi Zheng, Xinlei He†

USENIX Security USENIX Security Symposium
arXiv

GasAgent: A Multi-Agent Framework for Automated Gas Optimization in Smart Contracts

Jingyi Zheng*, Zifan Peng*, Yule Liu, Junfeng Wang, Yifan Liao, Wenhan Dong, Xinlei He†

arXiv arXiv preprint arXiv:2507.15761
arXiv

Cross-Chain Options: A Bridgeless, Universal, and Efficient Approach

Zifan Peng, Yingjie Xue†, Jingyu Liu

ICWS IEEE International Conference on Web Services
arXiv

Prompt-based Contrastive Learning to Combat the COVID-19 Infodemic

Zifan Peng, Mingchen Li, Yue Wang†, Daniel Y. Mo

MLJ Machine Learning, 114(1), 6
DOI
2024

Combating the COVID-19 Infodemic Using Prompt-Based Curriculum Learning

Zifan Peng, Mingchen Li, Yue Wang†, George T.S. Ho

ESWA Expert Systems with Applications, 229, 120501
DOI

Topic 01 · 9 papers

Trustworthy AI

Jailbreak robustness, AI-powered search safety, privacy risk, covert content, and reliable evaluation for frontier AI systems.

2026

What Your Posts Reveal: A Benchmark and Agentic Framework for User-Level Privacy Leakage on Social Media

Zifan Peng*, Yini Huang*, Aiwen Lu*, Qiming Ye, Peixian Zhang, Jingyi Zheng, Yule Liu, Xuechao Wang, Xinlei He†, Jiaheng Wei†

arXiv arXiv preprint arXiv:2606.06784
arXiv

Behind EvoMap: Characterizing a Self-Evolving Agent-to-Agent Collaboration Network

Qiming Ye*, Peixian Zhang*, Yupeng He, Zifan Peng, Gareth Tyson

arXiv arXiv preprint arXiv:2605.25815
arXiv

On the Generation and Mitigation of Harmful Geometry in Image-to-3D Models

Yule Liu, Yilong Yang, Jiale Teng, Hanze Jia, Zeren Luo, Jingyi Zheng, Zifan Peng, Ke Li, Yifan Liao, Zhen Sun, Jiaheng Wei, Yang Liu, Zhuo Ma, Xinlei He†

arXiv arXiv preprint arXiv:2605.09606
arXiv

JALMBench: Benchmarking Jailbreak Vulnerabilities in Audio Language Models

Zifan Peng, Yule Liu, Zhen Sun, Mingchen Li, Zeren Luo, Jingyi Zheng, Wenhan Dong†, Xinlei He†, Xuechao Wang, Yingjie Xue, Shengmin Xu, Xinyi Huang

ICLR International Conference on Learning Representations
arXiv
2025

Source Coverage and Citation Bias in LLM-based vs. Traditional Search Engines

Peixian Zhang*, Qiming Ye*, Zifan Peng, Kiran Garimella, Gareth Tyson

arXiv arXiv preprint arXiv:2512.09483
arXiv

Auditing Data Membership in Reinforcement Learning With Verifiable Rewards

Yule Liu, Heyi Zhang, Jinyi Zheng, Zhen Sun, Zifan Peng, Jiaheng Wei, Tianshuo Cong, Yilong Yang†, Xinlei He†

arXiv arXiv preprint arXiv:2511.14045
arXiv

CHASM: Unveiling Covert Advertisements on Chinese Social Media

Jingyi Zheng*, Tianyi Hu*, Yule Liu, Zhen Sun, Zongmin Zhang, Zifan Peng, Wenhan Dong†, Xinlei He†

NeurIPS Conference on Neural Information Processing Systems
Project

Unsafe LLM-Based Search: Quantitative Analysis and Mitigation of Safety Risks in AI Web Search

Zeren Luo*, Zifan Peng*, Yule Liu, Zhen Sun, Mingchen Li, Jingyi Zheng, Xinlei He†

USENIX Security USENIX Security Symposium
arXiv

Thought Manipulation: External Thought Can Be Efficient for Large Reasoning Models

Yule Liu, Jingyi Zheng, Zhen Sun, Zifan Peng, Wenhan Dong, Zeyang Sha, Shiwen Cui, Weiqiang Wang, Xinlei He†

arXiv arXiv preprint arXiv:2504.13626
arXiv

Topic 02 · 10 papers

Human-Centered AI

LLM evaluation, policy simulation, cognitive assessment, visual assistance, human-agent applications, and applied machine learning.

2026

"What Did It Actually Do?": Understanding Risk Awareness and Traceability for Computer-Use Agents

Zifan Peng, Mingchen Li

arXiv arXiv preprint arXiv:2603.28551
arXiv

GPLab: A Generative Agent-Based Framework for Policy Simulation and Evaluation

Shuhan Zhang†, Zifan Peng, Yinwang Ren

JASSS Journal of Artificial Societies & Social Simulation, 29(1), 1
DOI PDF

NL2Repo-Bench: Towards Long-Horizon Repository Generation Evaluation of Coding Agents

Zifan Peng, etc.

ICML International Conference on Machine Learning
ICML arXiv
2025

SuperGPQA: Scaling LLM Evaluation across 285 Graduate Disciplines

Zifan Peng, etc.

NeurIPS Conference on Neural Information Processing Systems
arXiv

ZPD-SCA: Unveiling the Blind Spots of LLMs in Assessing Students' Cognitive Abilities

Wenhan Dong, Zhen Sun, Yuemeng Zhao, Zifan Peng, Jun Wu, Jingyi Zheng, Yule Liu, Xinlei He†, Yu Wang, Ruiming Wang, Xinyi Huang, Lei Mo†

arXiv arXiv preprint arXiv:2508.14377
arXiv

"I Can See Forever!": Evaluating Real-time VideoLLMs for Assisting Individuals with Visual Impairments

Ziyi Zhang*, Zhen Sun*, Zongmin Zhang, Zifan Peng, Yuemeng Zhao, Zichun Wang, Zeren Luo, Ruiting Zuo, Xinlei He†

arXiv arXiv preprint arXiv:2505.04488
arXiv

Humanizing LLMs: A Survey of Psychological Measurements with Tools, Datasets, and Human-Agent Applications

Wenhan Dong*, Yuemeng Zhao*, Zhen Sun, Yule Liu, Zifan Peng, Jingyi Zheng, Zongmin Zhang, Ziyi Zhang, Jun Wu, Ruiming Wang, Shengmin Xu, Xinyi Huang, Xinlei He†

arXiv arXiv preprint arXiv:2505.00049
arXiv

Enhanced Structured Lasso Pruning with Class-wise Information

Xiang Liu*, Mingchen Li*, Xia Li†, Leigang Qu, Guansu Wang, Zifan Peng, Yijun Song, Zemin Liu, Linshan Jiang†, Jialin Li

arXiv arXiv preprint arXiv:2502.09125
arXiv

Prompt-based Contrastive Learning to Combat the COVID-19 Infodemic

Zifan Peng, Mingchen Li, Yue Wang†, Daniel Y. Mo

MLJ Machine Learning, 114(1), 6
DOI
2024

Combating the COVID-19 Infodemic Using Prompt-Based Curriculum Learning

Zifan Peng, Mingchen Li, Yue Wang†, George T.S. Ho

ESWA Expert Systems with Applications, 229, 120501
DOI

Topic 03 · 4 papers

DeFi & Blockchain

Secure and efficient decentralized protocols, smart contract analysis, gas optimization, payment channels, and cross-chain financial primitives.

2026

TxSum: User-Centered Ethereum Transaction Understanding with Micro-Level Semantic Grounding

Zifan Peng, Jingyi Zheng, Yule Liu, Huaiyu Jia, Qiming Ye, Jingyu Liu, Xufeng Yang, Mingchen Li, Qingyuan Gong, Xuechao Wang, Xinlei He†

arXiv arXiv preprint arXiv:2512.06933
arXiv
2025

GasAgent: A Multi-Agent Framework for Automated Gas Optimization in Smart Contracts

Jingyi Zheng*, Zifan Peng*, Yule Liu, Junfeng Wang, Yifan Liao, Wenhan Dong, Xinlei He†

arXiv arXiv preprint arXiv:2507.15761
arXiv

Cross-Chain Options: A Bridgeless, Universal, and Efficient Approach

Zifan Peng, Yingjie Xue†, Jingyu Liu

ICWS IEEE International Conference on Web Services
arXiv

FairRelay: Fair and Cost-Efficient Peer-to-Peer Content Delivery through Payment Channel Networks

Yingjie Xue†, Jingyu Liu, Zifan Peng, Chao Lin, Jianan Hong, Xinyi Huang

TDSC IEEE Transactions on Dependable and Secure Computing
Project