DivSkill-SQL Optimises Text-to-SQL Models with New Technology
A new research framework called DivSkill-SQL improves the performance of Text-to-SQL ensembles by optimising for complementary skills rather than individual models.
What happened?
Researchers have introduced DivSkill-SQL, a residual skill optimisation framework that creates complementary agentic Text-to-SQL ensembles without requiring model fine-tuning. Instead of merely drawing SQL query candidates and selecting the best one, DivSkill-SQL focuses on optimising new skills based on the examples where the current ensemble fails. This contributes to increasing candidate diversity and reducing correlated errors.
Key facts
| Prestandaförbättring Snowflake (Spider2-Lite) | +11.1 poäng |
|---|---|
| Prestandaförbättring BigQuery (Spider2-Lite) | +8.3 poäng |
| Påverkade basmodeller | Opus-4.6, GPT-5.4 |
”We present DivSkill-SQL, a residual skill optimization framework that builds complementary agentic Text-to-SQL ensembles without model fine-tuning: each new skill is optimized on examples the current skill ensemble fails on, provably targeting its marginal contribution to Pass@K.”
”On Spider2-Lite, DivSkill-SQL improves selected accuracy by up to +11.1 points on Snowflake and +8.3 on BigQuery over the strongest ensemble baseline, with consistent gains across two base models (Opus-4.6 and GPT-5.4).”
Why it matters
Traditional Text-to-SQL ensembles are improved by generating multiple SQL candidates and selecting the most correct one. However, efficiency is limited by Pass@K – the probability that at least one of the K candidates is correct. Existing methods generate diversity heuristically, which often leads to candidate sets dominated by similar errors. DivSkill-SQL's methodology addresses this by systematically improving the model's ability to handle problems the ensemble misses, provably increasing its marginal contribution to Pass@K.
Who is affected?
Researchers and developers within Natural Language Processing (NLP), database administrators, and users of Text-to-SQL systems seeking improved accuracy and reliability in SQL generation are affected. Companies using LLMs for database interaction can benefit from increased performance and reduced errors.
What else you should know
DivSkill-SQL demonstrated significant improvements on benchmarks such as Spider2-Lite, with up to +11.1 points on Snowflake and +8.3 on BigQuery compared to existing ensembles. The method also showed consistent improvements across base models like Opus-4.6 and GPT-5.4, and optimised skills were transferred between different dialects without retraining.
Quick answers about this story
Vad har hänt?
När hände det?
Varför spelar det roll?
Vilka modeller berörs?
The link opens in a new window and leads to the publisher's own site.
Källan har spårats automatiskt från utgivaren via Aheadlines signalkedja.
Get similar news straight to your inbox
The reader's room
Send in a question or an addition. The newsroom reads everything before it's published and replies when relevant. No AI-generated text – just people.
Sign in to submit a comment or question.
Read the article through your role
- Decide whether this affects strategy over 6–12 months or is just noise.
- Discuss with leadership: do we own the right question or does ownership need to move?
- Ask: what risk are we taking by NOT acting on this this quarter?
Generated angle — not editorial analysis of "DivSkill-SQL Optimises Text-to-SQL Models with New Technolog"