In our LangChain Labs study with @Harvey
, we looked at how to measure efficiency across verifier designs. We benchmarked 5 setups against Sonnet per-criterion as the reference.

→ View original post on X — @langchain