Is there a way to bring some explainability to pipeline-builder LLM processing, like the explainability sidebar in AIP Logic ?
Context: I’m trying to process document to do some entity extraction. With this prompting, it works but not on all document and not always consistently. In order to understand what is missing in the prompting, I need to have clarity about where the models are failing in their reasoning.
How can I investigate the model reasoning in pipeline builder ?
