The story’s score and claims

Comments on No. 36 of 122 · Thinking Machines Lab · edition of Tuesday 17 November 2026

Putting Task Expertise into RL Achieves SOTA Performance on Text-to-SQL

Summary

A research team (UIUC Kang Lab) reports ReViSQL-K2.6, a single fine-tuned model trained with reinforcement learning on a cleaned training set, BIRD-Platinum, that reaches 88.55% on the expert-verified Arcwise-Plat-SQL benchmark and exceeds the 92.96% human proxy with 16-sample self-consistency.

Comments

Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules