12 papers
Purpose IR: Compiling User Purpose for Privacy-Preserving Agent Workflow State
Dongsheng Chen, Xiangyu Zhao, Lei Ma +2
LLM-agent workflows routinely transform rich observations into plans, summaries, memories, handoffs, and other reusable state. This creates a privacy boundary that task relevance a…
SaplingGuard: A Multidimensional-Profile-Aware Multi-Agent Guardrail for Developmentally Safe Adolescent-LLM Interaction
Jing Tan, Yifan Liu, Yi Lin +6
As adolescents increasingly use LLMs in everyday life, ensuring safe and developmentally appropriate responses has become essential. However, existing LLM guardrails primarily targ…
SameFact: The Same Safety Facts Lead to Different Responses Across Interfaces
Dongsheng Chen, Jiaxin Zhang, Lei Ma +2
Safety evaluations often ask whether a model recognizes that an action is unsafe, whereas agent evaluations ask what the model chooses to do. Using safety judgments as evidence abo…
Says Block, Still Acts: Why LLM Safety Judgments Fail to Govern Action in LLM Agents
Dongsheng Chen, Xiangyu Zhao, Xin Yao +1
Large language model agents can correctly judge that an action should be blocked while still preferring to take it. We ask why this judgment-action disconnect arises, and whether e…
OpenAgentFlow: Enabling System-Wide Safety Boundaries for Heterogeneous AI Agent Fleets
Dongsheng Chen, Xiangyu Zhao, Xin Yao +1
AI agents powered by large language models are evolving from isolated assistants into heterogeneous systems in which multiple agents, planners, tools, and execution backends operat…
"Allow" to Achieve, Over-Privileged Inadvertently: The Unintended Cost of Task-Completion-Driven Pop-up Decisions in Mobile GUI Agents
Dongsheng Chen, Yuxuan Li, Guanhua Chen +5
Mobile GUI agents routinely encounter system permission dialogs during task execution, yet their ability to grant only permissions that are necessary for the delegated task remains…
Modeling LLM Unlearning as an Asymmetric Two-Task Learning Problem
Zeguan Xiao, Siqing Li, Yong Wang +4
Machine unlearning for large language models (LLMs) aims to remove targeted knowledge while preserving general capability. In this paper, we recast LLM unlearning as an asymmetric…
BeSafe-Bench: Unveiling Behavioral Safety Risks of Situated Agents in Functional Environments
Yuxuan Li, Yi Lin, Peng Wang +2
The rapid evolution of Large Multimodal Models (LMMs) has enabled agents to perform complex digital and physical tasks, yet their deployment as autonomous decision-makers introduce…
ARGOS: Automated Functional Safety Requirement Synthesis for Embodied AI via Attribute-Guided Combinatorial Reasoning
Dongsheng Chen, Yuxuan Li, Yi Lin +6
Ensuring functional safety is essential for the deployment of Embodied AI in complex open-world environments. However, traditional Hazard Analysis and Risk Assessment (HARA) method…
Toward Automated Robustness Evaluation of Mathematical Reasoning
Yutao Hou, Zeguan Xiao, Fei Yu +6
Large Language Models (LLMs) have demonstrated remarkable capabilities in various reasoning-intensive tasks. However, these models exhibit unexpected brittleness, often failing on…
ImPart: Importance-Aware Delta-Sparsification for Improved Model Compression and Merging in LLMs
Yan Yang, Yixia Li, Hongru Wang +4
With the proliferation of task-specific large language models, delta compression has emerged as a method to mitigate the resource challenges of deploying numerous such models by ef…
SeqAR: Jailbreak LLMs with Sequential Auto-Generated Characters
Yan Yang, Zeguan Xiao, Xin Lu +5
The widespread applications of large language models (LLMs) have brought about concerns regarding their potential misuse. Although aligned with human preference data before release…