Researchers have developed SKT, a novel data synthesis pipeline designed to enhance the ability of language model agents to identify, apply, and coordinate skills. SKT generates skill-grounded tasks and verified trajectories by selecting skill configurations, synthesizing tasks through rule-based and agent-based verification, and retaining only successful skill-use trajectories. This process produced 4,000 task packages and 27,164 verified trajectories from 2,000 public skills. To evaluate skill-use performance, the team also constructed SkillEval, a new benchmark dataset. Experiments demonstrated that fine-tuning models on SKT-generated data consistently improves skill-use capabilities across various benchmarks and agent harnesses. AI
IMPACT This research introduces a method to improve AI agent's ability to use skills, potentially leading to more capable and versatile AI systems.
RANK_REASON The cluster contains an academic paper detailing a new method for training AI agents. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
