Agent Framework Benchmark

An open-source benchmark for comparative evaluation of agent frameworks, using 107 data engineering tasks to test LangGraph, CrewAI, and AutoGen. Evaluation metrics include task pass rate, latency, and token cost.