1 post
A 4B open-weights model, post-trained with SFT and RL, cut Postgres query latency 44.7% across 113 join-heavy benchmark queries using pg_hint_plan.
