Text-to-SQL parsing increasingly relies on high-quality Chain-of-Thought (CoT) data to train specialized models. However, eliciting accurate reasoning paths for complex queries remains a formidable challenge, causing existing methods to suffer from severe data underutilization and fragile free-form generation. To systematically resolve these issues, we propose MSI-SQL, a Multi-task Self-Iterative framework that enhances both data quantity and quality. To expand data scale, we introduce a verification-based self-iterative loop that progressively conquers difficult queries. To ensure reasoning stability, we design a masked SQL completion auxiliary task that serves as a structural anchor, explicitly constraining the CoT generation space. Coupled with a novel mask-augmented soft self-consistency mechanism during inference, evaluations on the BIRD and Spider benchmarks demonstrate that MSI-SQL establishes a new state-of-the-art for open-source Text-to-SQL parsers.

The choice of token vocabulary affects the performance of machine translation. This paper aims to figure out what is a good vocabulary and whether we can find the optimal vocabulary without trial training. To answer these questions, we first provide an alternative understanding of vocabulary from the perspective of information theory. It motivates us to formulate the quest of vocabularization – finding the best token dictionary with a proper size – as an optimal transport (OT) problem. We propose VOLT, a simple and efficient solution without trial training. Empirical results show that VOLT beats widely-used vocabularies in diverse scenarios, including WMT-14 English-German translation, TED bilingual translation, and TED multilingual translation. For example, VOLT achieves 70% vocabulary size reduction and 0.5 BLEU gain on English-German translation. Also, compared to BPE-search, VOLT reduces the search time from 384 GPU hours to 30 GPU hours on English-German translation. Codes are available at https://github.com/Jingjing-NLP/VOLT.