Jiaming Ji

On Robustness of Vision-Language-Action Model against Multi-Modal Perturbations

Mesa and Mask: A Benchmark for Detecting and Classifying Deceptive Behaviors in LLMs

Debate with Image: Detecting Deceptive Behaviors in Multimodal Large Language Models

Safe Reinforcement Learning with ADRC Lagrangian Method

Mitigating Deceptive Alignment via Self-Monitoring

SafeVLA: Towards Safety Alignment of Vision-Language-Action Model via Constrained Learning

Generative RLHF-V: Learning Principles from Multi-modal Human Preference

Safe RLHF-V: Safe Reinforcement Learning from Multi-modal Human Feedback

SAE-V: Interpreting Multimodal Models for Enhanced Alignment