1 paper
Yangjie Xu, Lujun Li, Lama Sleem +7
Skills are getting increasingly used in recent agentic frameworks, performing robustly with proprietary LLMs. Yet, their effectiveness for small and medium-sized open-source models…