Meanwhile, orgs are earlier in their evals usage, with around half of respondents running offline evals on test sets and just over one-third running online evals on production data.

→ View original post on X — @langchain,