AI safety researcher

Yue Zhao

University of Southern California

Google Scholar dblp OpenReview ORCID Homepage

Last 12 months9 papers

Evaluation and benchmarks 3Attacks 2Analysis and theory 1

AI safety papers from the last year

On this site: 15 papers

September 2026

Controlled Decoding Attacks on Black-Box LLMs
Sep 29AttacksarXiv ↗

June 2026

SpecAlign: Efficient Specification-Grounded Alignment of Large Language Models via Synthetic Data
Jun 15arXiv ↗

May 2026

FORTIS: Benchmarking Over-Privilege in Agent Skills
May 9Evaluation and benchmarksarXiv ↗

April 2026

Auditable Agents
Apr 7arXiv ↗
No Attacker Needed: Unintentional Cross-User Contamination in Shared-State LLM Agents
Apr 1Evaluation and benchmarksarXiv ↗

February 2026

Benchmarking Knowledge-Extraction Attack and Defense on Retrieval-Augmented Generation
Feb 10Evaluation and benchmarksarXiv ↗

January 2026

"Someone Hid It": Query-Agnostic Black-Box Attacks on LLM-Based Retrieval
Jan 30AttacksarXiv ↗
Defenses Against Prompt Attacks Learn Surface Heuristics
Jan 12Analysis and theoryarXiv ↗

November 2025

A Systematic Study of Model Extraction Attacks on Graph Foundation Models
Nov 14arXiv ↗

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