Add statistical rigor, multi-benchmark evaluation, and paper rewrite for VLDB submission#116

Draft

Krishnachaitanyakc wants to merge 43 commits intomainfrom

kc/local-changes

Commits on Feb 8, 2026

Update .gitignore for evaluation framework artifacts
Krishnachaitanyakc
committed
Add column-name-based alignment to result comparator
Krishnachaitanyakc
committed
Fix critical prompt bugs and enhance schema context
Krishnachaitanyakc
committed
Increase LLM max tokens from 1024 to 2048
Krishnachaitanyakc
committed
Add self-correction loop for SQL generation errors
Krishnachaitanyakc
committed
Clean up benchmark queries for accurate evaluation
Krishnachaitanyakc
committed
Add chain-of-thought reasoning module for SQL generation
Krishnachaitanyakc
committed
Add re-evaluation script and integrate self-correction into Phase 2
Krishnachaitanyakc
committed
Add core evaluation framework for text-to-SQL experiments
Krishnachaitanyakc
committed
Add benchmark schemas, examples, configs, and analysis tools
Krishnachaitanyakc
committed
Add subset column matching to result comparator
Krishnachaitanyakc
committed
Add column reorder tolerance and row-superset matching to comparator
Krishnachaitanyakc
committed
Add LIMIT, JOIN, and window function guidance to prompts
Krishnachaitanyakc
committed
Add execution-guided result refinement to self-corrector
Krishnachaitanyakc
committed
Add self-consistency voting module for SQL generation
Krishnachaitanyakc
committed
Integrate refinement and self-consistency into Phase 2 pipeline
Krishnachaitanyakc
committed
Add fuzzy column name matching for result comparison
Krishnachaitanyakc
committed
Disable execution-guided refinement (net negative impact)
Krishnachaitanyakc
committed
Fix LIMIT guidance and improve re-evaluation script
Krishnachaitanyakc
committed
Use SEMANTIC strategy consistently for large result sets
Krishnachaitanyakc
committed
Fix gold SQL in aggregation benchmark queries to match question scope
Krishnachaitanyakc
committed
Fix gold SQL in clickhouse_specific benchmark query to match question scope
Krishnachaitanyakc
committed
Fix gold SQL in complex_joins benchmark queries to match question scope
Krishnachaitanyakc
committed
Fix gold SQL in time_series benchmark queries to match question scope
Krishnachaitanyakc
committed
Remove unjustified LIMIT clauses from few-shot examples
Krishnachaitanyakc
committed
Add percentage/rounding prompt guidance and relax numeric tolerance
Krishnachaitanyakc
committed
Remove unjustified LIMIT clauses from gold SQL benchmark queries
Krishnachaitanyakc
committed
Add fuzzy column name matching to column reorder logic
Krishnachaitanyakc
committed
Add ClickHouse function guidance and table relationship hints
Krishnachaitanyakc
committed

Commits on Feb 9, 2026