The field of Text-to-SQL is moving towards more structured and reliable approaches, with a focus on principled frameworks and software-engineering-inspired methods. Researchers are exploring ways to improve the accuracy and interpretability of SQL generation, such as using weighted consensus tournaments and declarative techniques to handle heterogeneous data sources. Another key area of development is the integration of natural language understanding and structured data access, with a focus on bridging the gap between these two areas. Noteworthy papers in this area include JudgeSQL, which introduces a reasoning-based SQL judge model, and DeepEye-SQL, which reframes Text-to-SQL as a software-engineering problem. OG-Rank is also notable for its single-decoder approach to ranking, which pairs a pooled first-token scoring signal with an uncertainty-gated explanation step.