[FR]: Bar chart (categorical) score metric e.g. "low", "medium", "high"

[gustavhartz]

Proposal summary

Make it possible for evaluation metrics that have categorical/text values

Motivation

What problem are you trying to solve: key approach within the company I currently work at is evaluation of LLM's with categorical rankings: e.g. low, medium, high, level of hallucination

**How are you currently solving this problem?:**Currently I generate plots based on the data manually after a run

What are the benefits of this feature?: This would make evaluation of model performance a lot easier

[aadereiko]

Hey @gustavhartz! A great idea! Thanks for reporting it. I am going to create a ticket internally so our team can take it from there.

OPIK-1150 - for our tracking

[gustavhartz]

@aadereiko sounds good! Plotting experiment works reasonably well with Weights & Biases, so might be a bit of inspiration to be found there.