Full threadjacky2wong·In this documentation for Confident AI, I outline a few weaknesses to BertScore and why it needs to be coupled with entailment to be a useful testing component for LLMs.View on HN