Скачать или смотреть Preview of Strategies for Testing AI-Based Agents: Ensuring Quality... | R. Sawant | STAREAST 2026

Strategies for Testing AI-Based Agents: Ensuring Quality and Reliability in Intelligent Systems
In an age increasingly shaped by AI-based agents, ensuring their reliability, safety, and performance introduces complex testing challenges. Traditional testing methods, designed for predictable and deterministic software, often fall short when applied to the adaptive, emergent, and opaque nature of AI systems. This presentation offers a forward-looking framework to address these challenges effectively. It will introduce foundational protocols such as the Model Context Protocol (MCP), Agent-to-Agent (A2A) Protocol, and Agent-to-GUI (AGUI) Protocol, which provide structured approaches for defining and validating agent interactions. The session will also explore advanced strategies, including Context Engineering to create diverse and dynamic test scenarios, the use of Large Language Models (LLMs) as evaluators—both as "LLM as a Judge" for automated assessments and "LLM as a Committee" for enhancing robustness—and the integration of Human-in-the-Loop (HITL) testing to handle complex edge cases and ethical considerations. Key takeaways for attendees will be practical insights and actionable techniques for incorporating these innovative approaches into AI-Agent testing workflows, equipping them to ensure the quality, fairness, and reliability of AI-based agents in an evolving technological landscape
https://stareast.techwell.com/program...