We need to do some qualitative analysis of the ask forge sessions and feedback that's been collated. Form a view of what needs to be built based on the feedback and sessions and build a dataset that will help evaluate how much the sessions are adhering to that view.
Once we've formed this 'view' or 'taste', we should extend the visualizer to help annotate the sessions in a manner that is helpful to build an eval dataset that align with the taste quickly.
We need to do some qualitative analysis of the ask forge sessions and feedback that's been collated. Form a view of what needs to be built based on the feedback and sessions and build a dataset that will help evaluate how much the sessions are adhering to that view.
Once we've formed this 'view' or 'taste', we should extend the visualizer to help annotate the sessions in a manner that is helpful to build an eval dataset that align with the taste quickly.