Xuan Zhang

This paper analyzes 470 security advisories in the OpenClaw AI agent framework, demonstrating that the system's structural weakness lies in per-layer trust enforcement, enabling cross-layer remote code execution (RCE) attacks.

TraceScope is an interactive, sandboxed triage pipeline that analyzes complex phishing URLs by simulating human interaction and verifying suspicious behavior against a detailed checklist, achieving high detection rates even against advanced, evasive threats.

This paper comparatively evaluates DKnownAI Guard against three competitors, demonstrating that DKnownAI Guard achieves superior performance in detecting both agent-specific threats and harmful content.

MIRA proposes a novel source-aware filtering framework that discovers and anchors evaluation rubrics during data selection, significantly improving code-oriented mid-training data quality while reducing token usage.

iLoRA introduces a novel Bayesian graph-conditioned LoRA framework that jointly learns prediction and latent interaction structure, significantly improving microbiome diagnosis by modeling microbe-microbe cross-talk.

The paper identifies and demonstrates that post-conclusion continuation in answer-correct long-CoT traces is harmful during LLM fine-tuning, proposing a method to cut this continuation.

The paper introduces PrivacyPeek, a new benchmark that audits the acquisition stage of LLM-based agents to show that unnecessary and sensitive data acquisition is a widespread and critical privacy vulnerability.

The paper introduces PrivacyPeek, a new benchmark that audits the acquisition stage of LLM-based agents to demonstrate that unnecessary acquisition of sensitive data is a widespread and critical privacy vulnerability.

The paper introduces Dr. DocBench, a difficulty-aware, comprehensive benchmark designed to rigorously test expert-level and challenging document parsing capabilities for VLMs, demonstrating that current state-of-the-art models fail on complex, domain-specific structures.

The paper introduces MindClaw, a closed-loop framework that enables embodied agents to perform real-time mental-state reasoning and intervene with precision, significantly outperforming standard VLM baselines.

This paper demonstrates that fusing multi-viewpoint data from multiple satellites significantly enhances the accuracy of space object detection in congested LEO constellations, establishing multi-view fusion as an effective strategy.

QUBRIC introduces a co-design framework that simultaneously optimizes queries and rubrics, overcoming the bottleneck of vague rubrics derived from open-ended questions, leading to significant gains in RL performance.

This paper proposes CompRank, a token-efficient reranking framework for large language models that reduces redundant computation and achieves strong reranking performance.