Thread

Beating GPT-5.6 Sol on retrieval with 100x cheaper open models

414 points · 114 comments · moonikakiss

punnerud · 1786032421 · focus · HN ↗
What’s in that “Search tool”, could actually be the same solution presented in a different way
1. kumama · 1786033431 · focus · HN ↗
  castform founder here. it uses lakebases's native bm25 and vector search and fuses the results using rrf (<a href="https://medium.com/@devalshah1619/mathematical-intuition-behind-reciprocal-rank-fusion-rrf-explained-in-2-mins-002df0cc5e2a" rel="nofollow">https://medium.com/@devalshah1619/mathematical-intuition-beh...)
2. softwaredoug · 1786034186 · focus · HN ↗
  Usually in these cases you don't need to do much tuning to the retriever. So you just give it BM25 or somesuch.
  I'm hesitant to say absolutely zero tuning, because there are cases where you do want to say, bias towards trustworthy results or recent results etc to help the model avoid wasting tokens. But probably not much beyond that.
  You can also just create a param in the tool for the agent that selects for "recent" or "popular" or "trustworthy" in ranking.

Unofficial Hacker News client; not affiliated with Y Combinator.