7 papers
SkillSentry: Adaptive Honey Worlds for Dynamic Safety Testing of Agent Skills
Nizhang Li, Zonghao Ying, Xiangfan Wu +7
External skills extend the capabilities of large language model agents, but also introduce an execution-time attack surface: a skill that appears benign under inspection may reveal…
SafeFlow: Semantic Information-Flow Control for Blocking Malicious Propagation in Multi-Agent Systems
Haowen Dai, Zonghao Ying, Wenfeng Li +10
Multi-agent systems improve capability through task decomposition and role specialization, but these same mechanisms introduce an important safety blind spot: a harmful objective c…
TrajShield: Trajectory-Level Safety Mediation for Defending Text-to-Video Models Against Jailbreak Attacks
Quanchen Zou, Nizhang Li, Wenxin Zhang +4
Text-to-Video (T2V) models have demonstrated remarkable capability in generating temporally coherent videos from natural language prompts, yet they also risk producing unsafe conte…
Placing Puzzle Pieces Where They Matter: A Question Augmentation Framework for Reinforcement Learning
Yangyi Fang, Jiaye Lin, Xiaoliang Fu +2
Reinforcement learning has become a powerful approach for enhancing large language model reasoning, but faces a fundamental dilemma: training on easy problems can cause overfitting…
From to : Taming Divergence in Soft Clipping via Bilateral Decoupled Decay of Probability Gradient Weight
Xiaoliang Fu, Jiaye Lin, Yangyi Fang +6
Reinforcement Learning with Verifiable Rewards (RLVR) has catalyzed a leap in Large Language Model (LLM) reasoning, yet its optimization dynamics remain fragile. Standard algorithm…
Tracing and Mitigating Hallucinations in Multimodal LLMs via Dynamic Attention Localization
Tiancheng Yang, Lin Zhang, Jiaye Lin +3
Multimodal Large Language Models (MLLMs) achieve strong performance on tasks like image captioning and visual question answering, but remain prone to hallucinations, where generate…
Curriculum-RLAIF: Curriculum Alignment with Reinforcement Learning from AI Feedback
Jiaye Lin, Mengdi Li, Xufeng Zhao +4
Reward models trained through Reinforcement Learning from AI Feedback (RLAIF) methods frequently suffer from limited generalizability, which hinders the alignment performance of po…