Lin Yan

The paper introduces TraceSafe-Bench, a comprehensive benchmark, and finds that securing LLM agents requires jointly optimizing for structural reasoning and safety alignment to mitigate risks during multi-step tool-use.

The paper proposes SRTJ, a Self-Evolving Rule-Driven Training-Free Jailbreak framework that systematically discovers and refines attack strategies using rule composition and feedback to achieve robust and generalizable jailbreaking against modern LLMs.

AgentTrust is a novel runtime safety layer that intercepts and evaluates AI agent tool calls before execution, achieving high accuracy in detecting unsafe actions across complex and obfuscated scenarios.

The paper introduces POLARIS, a novel framework that systematically generates comprehensive and verifiable safety tests for LLMs by formalizing natural language policies into First-Order Logic and exploring the resulting Semantic Policy Graph.

The paper introduces BITE, a black-box adversarial framework that exploits stylistic biases in LLM judges by adaptively generating semantically equivalent edits to artificially inflate assigned scores.

MOOSE-Copilot is a novel web-based framework that unifies scientific hypothesis discovery by formalizing human-AI interaction, significantly improving performance over autonomous LLM baselines.

The paper introduces PARL, a framework that learns personalized evaluation rubrics directly from raw user interaction histories to accurately assess how well LLM outputs align with subjective, user-specific preferences.

This paper introduces personalized empathy, a capability for LLMs to adapt empathetic strategies based on individual user history, and proposes PereGRM, a reward modeling framework that significantly enhances this personalized empathy.

This paper introduces a machine learning model, RuBR, and a methodology to reliably distinguish genuine astronomical transients from spurious detections for the upcoming Roman Space Telescope's data pipeline.