Evaluations is usually quantitative, which can result in info loss, or qualitative, leveraging the semantic strengths of LLMs to keep multifaceted information. In place of manually creating them, you could consider to leverage the LLM by itself to formulate possible rationales for the forthcoming move.Generalized models might have equal efficiency

If a basic prompt doesn’t yield a satisfactory response through the LLMs, we must always offer the LLMs specific Guidance.This “chain of imagined”, characterised with the pattern “dilemma ? intermediate issue ? comply with-up thoughts ? intermediate question ? stick to-up inquiries ? … ? final remedy”, guides the LLM to reach the ultima