Target
Generate or select executable SQL on frozen schemas.
Failure this recipe addresses
Small-model SQL gains fail to generalize across public and synthetic domains or emit prose wrappers.
Data contract
Historical frozen SQL fixtures, public-style rows, five synthetic SQLite domains, and preference pairs.
Method or policy
SFT, routing, composition, and DPO were compared under their recorded gates; remaining proposals are closed rather than queued.
Evaluation contract
SQLite execution plus exact and slice metrics.
- joins
- filters
- aggregates
- format
- public
- synthetic
Budget and stop rule
Historical lineage closed; no additional compute is authorized by this registry.
The lineage stopped when stronger DPO pressure still could not change the output-format prior and no new target was frozen.
Decision rule
Retain the routed SQL candidate and failed attempts as learning evidence; start future SQL work from a fresh question.
Learning exercise
Reconstruct the SQL lineage and mark which variable changed at each step.
Explain why routing beat blending and why DPO failed to remove the wrapper.
How to read this dossier
This canonical page separates the retained repository decision from the material that informed it. A source may describe an external claim, historical measurement, local observation, or planned procedure; those evidence classes are not interchangeable. Follow the provenance links before reusing a number or method.
Status and disposition describe what PostTrainLLM retained when this record was indexed. They are not a live product promise, a newly run benchmark, or permission to restart historical work. Unknown or unmeasured fields remain unknown rather than being treated as zero.
To reuse the record, first name the exact claim you need and trace it to the linked source. Then check whether the original environment, model revision, data split, evaluator, hardware, and budget match the proposed use. If they do not, treat the record as a hypothesis or design reference and run the smallest fresh comparison that can falsify it. Preserve negative outcomes and regressions beside any improvement; a local win on one slice does not silently become a general capability claim.
Source provenance
The normalized record comes from docs/recipes/registry.json. The links below are the tracked evidence and explanatory sources preserved with the record.