SQLConductor uses Search-to-Policy Learning with MCTS, stability-weighted SFT, and curriculum RL to train a compact policy for adaptive step-wise Text-to-SQL orchestration, reporting 73.2% EX on BIRD-Dev.
arXiv:2606.13707 [cs.AI] https://arxiv.org/abs/2606.13707
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DB 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SQLConductor: Search-to-Policy Learning for Step-wise Text-to-SQL Orchestration
SQLConductor uses Search-to-Policy Learning with MCTS, stability-weighted SFT, and curriculum RL to train a compact policy for adaptive step-wise Text-to-SQL orchestration, reporting 73.2% EX on BIRD-Dev.