Kai Wang
5 papers · Latest:
Good Agentic Friends Do Not Just Give Verbal Advice: They Can Update Your Weights
TFlow enables multi-agent LLMs to communicate via transient weight perturbations, boosting efficiency and accuracy over text-based methods.
SkillSafetyBench: Evaluating Agent Safety under Skill-Facing Attack Surfaces
SkillSafetyBench evaluates how reusable skills in LLM agents create new attack surfaces, revealing vulnerabilities beyond model-level alignment.
Audio-Visual Intelligence in Large Foundation Models
This survey provides the first comprehensive review of Audio-Visual Intelligence (AVI) in large foundation models, unifying tasks, methods, and challenges.
The Salami Slicing Threat: Exploiting Cumulative Risks in LLM Systems
This paper introduces Salami Slicing, a novel multi-turn jailbreak attack that exploits cumulative low-risk inputs to bypass LLM safety, achieving high success rates.
Beyond Loss Values: Robust Dynamic Pruning via Loss Trajectory Alignment
AlignPrune robustly prunes data under noisy labels by using loss trajectory alignment, outperforming existing dynamic pruning methods.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.