Uploaded December 2025 | Updated September 2026, 3 weeks ago
Evaluating generative AI is hard because “good” can be objective (facts) or wildly subjective (style, preference, persona). In this video we split the problem into two parts: formatting and clarity and user-relative quality. Learn show how to validate an LLM-judge (treat it like any other model: label, test, iterate), and how to attach personality or persona to judges so evaluations match user needs. If you build recommender, writing, or conversational systems, this walkthrough gives practical evaluation patterns you can adopt today; plus pitfalls to avoid when trusting an LLM as the arbiter.
#GenAI #LLM #Evaluation #AIAlignment #PromptEngineering #NLP #MachineLearning #HumanCenteredAI #AIEthics #Personalization
Evaluating generative AI is hard because “good” can be objective (facts) or wildly subjective (style, preference, persona). In this video we split the problem into two parts: formatting and clarity and user-relative quality. Learn show how to validate an LLM-judge (treat it like any other model: label, test, iterate), and how to attach personality or persona to judges so evaluations match user needs. If you build recommender, writing, or conversational systems, this walkthrough gives practical evaluation patterns you can adopt today; plus pitfalls to avoid when trusting an LLM as the arbiter.
#GenAI #LLM #Evaluation #AIAlignment #PromptEngineering #NLP #MachineLearning #HumanCenteredAI #AIEthics #Personalization
