TRACE: Improved Credit Assignment for Multi-Turn Jailbreaking
Discover TRACE, a novel framework enhancing credit assignment in multi-turn jailbreaking, boosting attack success and defense in large language models.
Mixed-Policy Distillation for Efficient AI Reasoning
Discover how Mixed-Policy Distillation compresses reasoning in AI models, reducing token use and boosting efficiency in large language models.
EvoMAS: Adaptive Workflows for Multi-Agent Systems
Discover EvoMAS, a novel framework that enables real-time adaptive workflows to optimize multi-agent systems in dynamic environments.
EDMolGPT: GPT-Style Drug Design Using Electron Density
Discover EDMolGPT, a novel GPT-based framework leveraging electron density for accurate, flexible structure-based drug design and faster discovery.
AHD Agent: Reinforcement Learning for Smart Heuristic Design
Discover how AHD Agent uses agentic reinforcement learning to enhance automatic heuristic design for complex optimization problems efficiently.