Hello! I'm trying to make sense of the dataset "reference" field. I'm a little confused as to how I should leverage it.
Should this be an empty column in my csv to place experiment output?
What does the example in the help text actually mean? I don't understand why dataset.reference maps to dataset.attributes.output.values
This is a mapping, typically used to represent the output column. It can be used as a shortcut when writing eval templates to compare against experiment outputs. For example {dataset.reference} maps to {dataset.attributes.output.value}
Thanks
Hi Brandon, the RunLLM answer is correct. It's optional - but teams use it so they can quickly reference it in their eval templates. This is helpful esp if you have a dataset column called {output} and you want to compare it to your experiment {output}. It helps differentiate between the two different outputs when the dataset one can be referred to as reference.
Thanks for confirming Aparna, the Agent was helpful.
