FORTIS Benchmark: Detecting Over-Privilege in AI Skills
Discover how the FORTIS benchmark reveals over-privilege in AI agent skills, highlighting risks and guiding safer, more precise model behavior.
Enhancing LLM Reasoning with Dynamic Persona Polylogues
Explore how dynamic persona vectors improve reasoning and accuracy in large language models through real-time monitoring and intervention.
BoostAPR: Advanced Reinforcement Learning for Program Repair
Discover BoostAPR, a cutting-edge framework using dual reward models and reinforcement learning to revolutionize automated program repair accuracy.
MCP-Cosmos: Enhancing Task Execution with World Models
Discover how MCP-Cosmos integrates World Models with MCP to boost agent task execution and predictive automation in complex environments.
Data-driven Circuit Discovery for Interpreting Language Models
Discover how Data-driven Circuit Discovery (DCD) enhances interpretability of language models by revealing multiple task-specific circuits.