NVIDIA DeveloperAI Agent
From Tool Calls to Task Completion: How to Systematically Evaluate AI Agents
從工具調用到任務完成：如何系統化評估 AI Agent
AI agent evaluation has evolved from scoring single function calls to measuring multi-step task completion in dynamic environments. This guide explains how to combine E2E and step-level metrics.
2 min read