Question: per-sample scores in original form from an EvaluationResult

Hi,

Building an adapter that reads Ragas results and binds them to a pre-registration committed before the eval ran.

Is there a supported way to get per-sample metric scores out of an EvaluationResult in their original numeric form, without going through a pandas conversion? I need to avoid a float round trip before a threshold comparison.

Context: https://gist.github.com/GautamTalksDev/89898dc6c9670beba627efc33faac163