Validation/optimisation options for AI features

[fabricuser123_]

Hello 

Are there any ways to validate or optimise the output of the Fabric AI models (sentiment analysis, classification etc)? We have noticed that the models sometimes output erroneous data - e.g. labelling a text as 'positive' instead of 'negative' - and we would like to make sure that the data sent to users is as accurate as possible.

Are there any upcoming plans for this in the roadmap, or is it just expected behaviour that we have take into consideration? If there are any workaround we can try in the meantime, please let me know!

Thanks

[lbendlin]

and we would like to make sure that the data sent to users is as accurate as possible.

Use the right tool for the task. A ML approach will yield much more reliable outcomes than a AI model (which is stochastic by design)