Skills-Coach: A Self-Evolving Skill Optimizer via Training-Free GRPO

TL;DR

Skills-Coach is an innovative framework that self-evolves and optimizes skills in LLM-based agents through training-free modules, improving performance across diverse skill categories.

Contribution

It introduces a comprehensive, training-free skill optimization framework with modular components for skill generation, optimization, evaluation, and benchmarking.

Findings

Achieves significant performance improvements across various skills.

Introduces Skill-X, a benchmark dataset with 48 diverse skills.

Demonstrates flexible execution modes for practical deployment.

Abstract

We introduce Skills-Coach, a novel automated framework designed to significantly enhance the self-evolution of skills within Large Language Model (LLM)-based agents. Addressing the current fragmentation of the skill ecosystem, Skills-Coach explores the boundaries of skill capabilities, thereby facilitating the comprehensive competency coverage essential for intelligent applications. The framework comprises four core modules: a Diverse Task Generation Module that systematically creates a comprehensive test suite for various skills; a Lightweight Optimization Module dedicated to optimizing skill prompts and their corresponding code; a Comparative Execution Module facilitating the execution and evaluation of both original and optimized skills; and a Traceable Evaluation Module, which rigorously evaluates performance against specified criteria. Skills-Coach offers flexible execution options through its virtual and real modes. To validate its efficacy, we introduce Skill-X, a comprehensive benchmark dataset consisting of 48 diverse skills. Experimental results demonstrate that Skills-Coach achieves significant performance improvements in skill capability across a wide range of categories, highlighting its potential to advance the development of more robust and adaptable LLM-based agents.