10 papers
Benchmarking Jailbreak Guardrails for Embodied Agents
Xunguang Wang, Qingyue Wang, Yuguang Zhou +3
Embodied agents powered by large language models and vision-language models are increasingly deployed in physical environments, but jailbreak attacks can induce these agents to per…
Decompiling Quantum Assembly into Structured Programs
Pingchuan Ma, Zhantong Xue, Zongjie Li +6
Quantum compilers translate programs into native gates, insert SWAP gates to route interactions onto a device, and optimize the result. The output is quantum assembly, a flat gate…
Well-Typed ComfyUI Cannot Go Wrong: Gradual Refinement Types for Generative-AI Workflow Graphs
Pingchuan Ma, Zhantong Xue, Hui Li +4
Generative-AI workflows are programs assembled from independently developed models and operators. Their correctness depends on relationships among model families, tensor geometry,…
Sound Debloating of Redundant Checks in Zero-Knowledge Machine-Learning Circuits
Zhantong Xue, Pingchuan Ma, Zhaoyu Wang +3
Zero-knowledge (ZK) proof systems for neural-network inference compile the model into a system of arithmetic constraints. Many of these constraints are redundant checks: range proo…
AutoSpec: Safety Rule Evolution for LLM Agents via Inductive Logic Programming
Pingchuan Ma, Zhaoyu Wang, Zimo Ji +5
Large language model (LLM) agents increasingly automate complex tasks by integrating language models with external tools and environments. However, their autonomy poses significant…
From Shield to Target: Denial-of-Service Attacks on LLM-Based Agent Guardrails
Yuguang Zhou, Xunguang Wang, Pingchuan Ma +3
LLM-based guardrails have emerged as a highly effective defense against prompt injection and jailbreak attacks in autonomous agents. However, we reveal that the very reasoning and…
ZK-Value: A Practical Zero-Knowledge System for Verifiable Data Valuation
Zhaoyu Wang, Pingchuan Ma, Zhantong Xue +4
Data valuation is a foundational task in data marketplaces, where a Shapley-value attribution determines how a buyer's payment is distributed among data providers. Typically, the m…
Beyond Content Safety: Real-Time Monitoring for Reasoning Vulnerabilities in Large Language Models
Xunguang Wang, Yuguang Zhou, Qingyue Wang +5
Large language models increasingly rely on explicit chain-of-thought reasoning to solve complex tasks, yet the safety of the reasoning process itself remains largely unaddressed. E…
RepoMark: A Data-Usage Auditing Framework for Code Large Language Models
Wenjie Qu, Yuguang Zhou, Bo Wang +4
The rapid development of Large Language Models (LLMs) for code generation has transformed software development by automating coding tasks with unprecedented efficiency. However, th…
Prompt Inversion Attack against Collaborative Inference of Large Language Models
Wenjie Qu, Yuguang Zhou, Yongji Wu +4
Large language models (LLMs) have been widely applied for their remarkable capability of content generation. However, the practical use of open-source LLMs is hindered by high reso…