Clarification of which responses were the human labels based on

[yisz]

Can you please clarify for how to match responses <> human labels?

Human labels have an ID / Dataset / Model used, but there are multiple Response files given the same ID / Dataset / Model, varied by retrieved contexts, prompts etc.

(Thought it was the validation_small partition, dpr retrieval, but the indicies don't match)