Harnessing Splits in your Dataset with Arize Phoenix

Arize Phoenix now lets you run super-targeted experiments over specific slices of your dataset using Splits — like hard examples, train/val/test, or any other category you care about. In this video, we walk through how to set up splits in Arize Phoenix, wire them into your code, and use them to run focused experiments on your LLM triage assistant. Timestamps 00:00 – Why splits matter for LLM evaluation 01:00 – Support ticket triage dataset overview 03:35 – Creating and assigning splits in the Phoenix UI 05:40 – Running focused experiments over a split via code 08:15 – Using splits in the Prompt Playground & wrap-up More about splits for dataset experiments in AI agent development: https://arize.com/docs/phoenix/datasets-and-experiments/how-to-experiments/splits