2 Comments
User's avatar
AI & Ops Essentials's avatar

Thank you for the blog , very interesting read and that video link was really helpful to understand the context window aspect in detail.

Would also be interesting if there is a separate detailed article about Quality metrics .

Kannan Kalidasan's avatar

Sure, definitely great ask and noted! Thank you for the feedback.

When it comes to data, we have well-defined frameworks for measuring quality through dimensions like accuracy, integrity, uniqueness, consistency, and completeness.

Similarly, AI systems should be evaluated using a set of quality metrics. Many of these naturally build on the evaluation metrics in machine learning, but LLM-powered systems would introduce additional considerations.

That extend to capture unique characteristics of LLM-based systems, such as groundedness, faithfulness, tool usage, retrieval quality, and agent outcomes.