Reproducible recipe

Constrained generation

Structured tool calls, SQL, and schema-bound outputs. A capable model emits invalid syntax or forbidden output shapes.

Status
validated with caveat
Learning path
runtime and agents
Regression slices
4
Execution
Fresh experiment required

Target

Structured tool calls, SQL, and schema-bound outputs

Failure this recipe addresses

A capable model emits invalid syntax or forbidden output shapes.

Data contract

The target's existing frozen eval; no training data is required for the inference policy.

Method or policy

Restrict decoding to the grammar or legal token set and score raw validity separately from task correctness.

Evaluation contract

Schema or grammar validity plus the target semantic metric.

  • raw unconstrained accuracy
  • abstention
  • latency
  • unsafe action rate

Budget and stop rule

No training; one bounded baseline/candidate evaluation.

Stop if validity rises but semantic correctness, safety, or latency misses its frozen gate.

Decision rule

Ship only when constrained execution passes both validity and semantic gates; otherwise retain as a diagnostic.

Learning exercise

Run a tiny legal-token or JSON-schema fixture and compare raw versus constrained validity.

Explain why legal output is not the same as correct output.

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.