Target
Adapt a frozen Mac-local base to one narrow behavior.
Failure this recipe addresses
The base misses a bounded task and full fine-tuning is unnecessary or too costly.
Data contract
Frozen train/validation/test splits with a tiny-overfit gate and breadth data.
Method or policy
Train named target modules at a fixed rank and inspect effective-update geometry before changing capacity.
Evaluation contract
Target before/after score on a sealed holdout.
- breadth
- safety
- format
- adapter geometry
- serve/load parity
Budget and stop rule
Fresh experiment only; one tiny-overfit and one bounded candidate before rank or module sweeps.
Stop if tiny overfit fails, the target plateaus, or required breadth regresses beyond the frozen threshold.
Decision rule
Ship, route, retry, or reject from the same frozen report card.
Learning exercise
Run or inspect LoRA geometry for one successful and one failed adapter.
Explain rank, target modules, effective rank, and why lower loss is insufficient.
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.