HyDE

Hard200 pts0 solves
Instead of embedding the question, you ask the LLM to generate a hypothetical answer, then embed THAT for search. What gets embedded? Flag format: CONGRESS{embed_[what]} Example: CONGRESS{embed_question_keywords}
Hint
The hypothetical answer is closer in embedding space to real answers than the question is.