Adaptive Table Retrieval Method for Text-to-SQL
A new adaptive table retrieval method for text-to-SQL tasks adjusts the number of tables retrieved per query, overcoming the limitations of fixed-k approaches. The method uses an adaptive thresholding mechanism and a sliding-window reranking algorithm to efficiently process large table corpora. Experiments demonstrate its effectiveness.
Key facts
- Proposed adaptive table retrieval method adjusts number of tables per query.
- Uses adaptive thresholding mechanism for selective retrieval.
- Integrates sliding-window reranking algorithm.
- Addresses limitations of fixed-k retrieval approaches.
- Designed for text-to-SQL tasks.
- Published on arXiv with ID 2605.18766.
- Announce type: cross.
Entities
Institutions
- arXiv