SQL generation has finally reached the 80% threshold, non-technical people can now directly ask the database, and this integration with BigQuery is very practical.

View Original
CoinNetwork
Google releases Gemini-SQL2: tops the BIRD single-model leaderboard with 80.04% accuracy
Google Research Team Announces Gemini-SQL2, Based on Gemini 3.1 Pro, Achieving a Record 80.04% Execution Accuracy on the BIRD Single Model Leaderboard.
Compared to Gemini-SQL 2.5 Pro's 77.14%, an improvement of 2.9 percentage points.
BIRD covers over 12,700 question-answer pairs, evaluating multi-table joins, dirty data handling, and complex reasoning, testing whether SQL can execute correctly in the database.
In the future, capabilities will be integrated into BigQuery Studio, AlloyDB AI, and Cloud SQL Studio to help non-technical users query enterprise databases using natural language.
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned