Apply Score Analytics to Dataset Runs

[Xixaus]

Describe the feature or potential improvement

The new Score Analytics with Multi-Score Comparison feature is great for analyzing traces. It would be really useful to have similar functionality for dataset runs.

This would enable comparing multiple metrics within a single run (e.g., Human annotations vs. LLM as a Judge) or comparing the same metric across different runs (e.g., GPT-4 vs. Claude, prompt v1 vs. v2). This would make dataset evaluations much more powerful for experimenting with prompts and models.

Additional information

No response

[marliessophie]

Thanks for sharing! We'll likely start working on this towards the end of the quarter. I will keep you posted here when we do.