X-AgentRed: Red-Teaming Personal Agents in Evolving Multi-Service Ecosystems
To address this, we propose X-AgentRed, a self-evolving red-teaming framework that discovers harmful tool call paths through a graph of tool dependencies and expands and refines them using execution feedback.
ProofPaper ↗
Key points
- Personal agents increasingly connect to services across the web and applications by using tools.
- Proactively identifying these compositional failures requires systematically exploring diverse cross-service paths.
- By analyzing each tool's effects, potential harms, and preconditions for execution, the red-teaming agent constructs a directed graph that captures how one tool's effects can satisfy another's execution preconditions.
- Across four frontier agentic models, X-AgentRed achieves $1.77\times$ the harmful path rate of STAC and explores a more diverse set of tools, yielding $2.99\times$ the tool coverage on average.
Sources (1)
- [1]X-AgentRed: Red-Teaming Personal Agents in Evolving Multi-Service EcosystemsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 05:34 PM
To address this, we propose X-AgentRed, a self-evolving red-teaming framework that discovers harmful tool call paths through a graph of tool dependencies and expands and refines them using execution feedback.
Personal agents increasingly connect to services across the web and applications by using tools.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 9, 2026QTestBench: A Test Framework for Evaluating LLM-Generated Quantum Code
- Oct 9, 2026Tracing and Taming Hallucinated Terms via Subsequence-Association Filtering and Breadth-Aware Selection for LLM Query Expansion
- Oct 9, 2026Qwen/Qwen-Image-2.1-Turbo
- Oct 9, 2026Surfing the Activation Wave: Analyzing the Information and Geometry of LLM Hidden States for Speculative Decoding
- Oct 8, 2026Microsoft Joins the Local AI Push
- Oct 8, 2026Understanding Jev, the new model everyone is talking about