Xiangzheng Zhang

SafeHarbor is a novel, hierarchical memory-augmented framework that establishes context-aware decision boundaries for LLM agents, achieving state-of-the-art safety while minimizing over-refusal.

This paper introduces Agentic Workflow Injection (AWI), a new class of vulnerability in LLM-powered GitHub Actions, and presents TaintAWI, a novel taint-analysis tool that identifies hundreds of exploitable zero-day vulnerabilities.

The paper proposes DMN, a compositional jailbreak framework that utilizes distributed instructions, multimodal evidence, and a number chain task across multiple images to significantly enhance the attack success rate against multimodal LLMs.

The paper introduces Harness-Bench, a diagnostic benchmark that measures how different system 'harnesses' affect LLM agent performance in realistic workflows, showing that agent capability must be reported at the model-harness configuration level.

This paper synthesizes over 150 scattered studies and reports to provide the first comprehensive primer on post-training reasoning data, organizing the field around data objects, utility, construction, and scalability.