Arize AI Community Icon

Home
Events

Arize AX Releases
Arize AX Support
Arize News
Arize Observe
Discussions
Introductions
Phoenix Support

Powered by Tightknit

Best Practices for Naming Variables in Playground Evaluators | Arize AI Community

Arize AX Support

Arize AX Support

Best Practices for Naming Variables in Playground Evaluators

·Dec 19, 2025 05:09 PM·

Hello - When adding an evaluator to a playground, what is the variable name we should use in the eval prompt to reference the "input"? My definition of the input would be the prompt injected with variable values.

37 comments

· Sorted by Oldest

Kyle J.
·
·
Kyle J.
·
·
experiment
Kyle J.
·
·
yes
Kyle J.
·
·
and my next question that might impact this...
Kyle J.
·
·
i'm trying to create a dataset i can use for prompt optimization
Kyle J.
·
·
i was hoping to create the llm output and the eval explanation in the experiment to then go into a dataset
Kyle J.
·
·
so experiment > dataset > prompt optimization
Kyle J.
·
·
is this the best way to do this?
Kyle J.
·
·
can i create the dataset within the UI that would generate llm output and eval explanation from a given list of inputs for the prompt?
Kyle J.
·
·
Well the issue is that I don't have all the individual input values that are injected into the prompt right now in the spans. So the "input prompt" between spans are different due to different values that are injected. So I was thinking that would impact the prompt optimizer. So I was thinking of doing an experiment with just a few input values. In the screenshot: "kilos"
Kyle J.
·
·
i'm just trying to test the prompt optimizer
Kyle J.
·
·
with fake data
Kyle J.
·
·
ok trying
Kyle J.
·
·
Kyle J.
·
·
even if i get this to work, i'm not sure how i'm going to be able to get the data from experiment into dataset