Wei Yang

The paper proposes SALO, a novel detector that monitors the dynamic, layer-wise activation pattern (Refusal Trajectory) to improve jailbreak detection robustness compared to traditional methods relying on static terminal representations.

AESOP introduces an adversarial attack that targets the entire execution path of deep learning pipelines, demonstrating that path-aware selection can inflate computational costs by orders of magnitude more than single-model attacks.

The paper introduces FraudBench, a multimodal benchmark designed to detect AI-generated fraudulent refund evidence, finding that current AI models struggle significantly with claim-conditioned fake-damage detection.

The paper systematically measures the risk of current image-to-3D models generating harmful geometries, finding that these models are effective at reconstruction and existing safeguards are insufficient.

The paper proposes a framework called RaMem to improve the validity of retrieved memories in long-term memory systems for LLM agents.

This paper tests the influence of Big Five personality on partner selection in multi-agent LLM systems, finding strong and task-stereotyped effects that diverge from human team performance.