AI safety researcher

Haoyu Wang

Huazhong University of Science and Technology

Google Scholar dblp ORCID

Last 12 months12 papers

Attacks 2Analysis and theory 1

AI safety papers from the last year

On this site: 39 papers

July 2026

Skills That Don't Exist: A Large-Scale Study of Hallucinated Skill Recommendation in LLM Agents
Jul 14AttacksarXiv ↗
Ghosts Beneath Textures: Texture-Relation Cues for Cross-Paradigm AI-Generated Image Detection
Jul 4arXiv ↗
AgentFlow: Building Agent Dependency Graphs for Static Analysis of Agent Programs
Jul 2arXiv ↗

June 2026

Acoda: Adversarial Code Obfuscation for Defending against LLM-based Analysis
Jun 10arXiv ↗

May 2026

Unsafe by Flow: Uncovering Bidirectional Data-Flow Risks in MCP Ecosystem
May 8Analysis and theoryarXiv ↗
Demystifying and Detecting Agentic Workflow Injection Vulnerabilities in GitHub Actions
May 8AttacksarXiv ↗

April 2026

February 2026

January 2026

An Effective and Cost-Efficient Agentic Framework for Ethereum Smart Contract Auditing
Jan 25arXiv ↗

December 2025

Beyond Correctness: Exposing LLM-generated Logical Flaws in Reasoning via Multi-step Automated Theorem Proving
Dec 29arXiv ↗

Authorship follows dblp and the author's own Google Scholar profile; only AI safety papers on this site are listed.