Evaluation UI: UX Inconsistency

Is your feature request related to a problem?

When uploading a new dataset, users provide a CSV with question and answer columns. These become the golden dataset's question and
ground truth.

However, in evaluation results, we display columns: question, answer, and ground_truth — where answer is now the
generated response while question and ground_truth are from the original upload. This naming inconsistency creates confusion from a UX
perspective.

Describe the solution you'd like

• Rename the CSV upload columns to question and ground_truth (instead of question and answer)
• This aligns the upload schema with the evaluation results display
• Consistent naming: question = input, ground_truth = expected answer, answer = generated response

Alternative

Keep upload columns as-is but rename ground_truth column in evaluation results UI to expected_answer to differentiate it more clearly
from answer (generated response).

Commencez par suivre le flux d’importation CSV et l’UI des résultats d’évaluation afin de trouver où les noms de colonnes sont définis. Examinez le pull request associé et confirmez quel choix de nommage a été implémenté ; le travail est considéré comme terminé lorsque le schéma d’importation et l’affichage des résultats utilisent des noms clairs et cohérents pour la question, la Ground Truth et la réponse générée.