Open zsimjee opened 5 months ago
I've updated the validate_each_sentence
and validate_full_text
methods in the ProvenanceV0
validator to include cosine similarity values in the validation metadata, enhancing validation process with quantitative analysis of text relevance. These changes allow users to access detailed similarity measures directly in the validator_logs
, improving transparency and facilitating deeper insights into the validation outcomes.
If this approach proves beneficial, should I consider applying similar updates to ProvenanceV1
as well?
https://github.com/guardrails-ai/guardrails/discussions/574