Xuandong Zhao

Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces

Learning to Reason without External Rewards

In-Context Watermarks for Large Language Models

AgentSynth: Scalable Task Generation for Generalist Computer-Use Agents

Dataset Protection via Watermarked Canaries in Retrieval-Augmented LLMs

InfoSynth: Information-Guided Benchmark Synthesis for LLMs

Are You Getting What You Pay For? Auditing Model Substitution in LLM APIs

PromptArmor: An Essential Baseline for Prompt Injection Defenses

Confidence-Guided MCTS for Efficient Long-Horizon Web Agent Tasks

Assessing Judging Bias in Large Reasoning Models: An Empirical Study

MMDT: Decoding the Trustworthiness and Safety of Multimodal Foundation Models

LeakAgent: RL-based Red-teaming Agent for LLM Privacy Leakage

An Undetectable Watermark for Generative Image Models

Permute-and-Flip: An optimally stable and watermarkable decoder for LLMs

Scalable Best-of-N Selection for Large Language Models via Self-Certainty

Multimodal Situational Safety

Weak-to-Strong Jailbreaking on Large Language Models

Improving LLM Safety Alignment with Dual-Objective Optimization

Weak-to-Strong Jailbreaking on Large Language Models

DIS-CO: Discovering Copyrighted Content in VLMs Training Data

ClinicalLab: Aligning Agents for Multi-Departmental Clinical Diagnostics in the Real World

Efficiently Identifying Watermarked Segments in Mixed-Source Texts