Target
Gradient-free optimization when rewards are evaluable but differentiation is unavailable.
Failure this recipe addresses
The objective cannot provide useful gradients or the model/policy is treated as a black box.
Data contract
A deterministic task set and scalar reward with repeated-seed variance estimates.
Method or policy
Sample parameter or prompt perturbations, evaluate them under the same seeds, and update toward higher reward.
Evaluation contract
Held-out reward against the unchanged incumbent.
- reward variance
- evaluation cost
- breadth
- seed sensitivity
Budget and stop rule
Not run in the closed project; any use is a fresh experiment with a fixed evaluation-call ceiling.
Stop when confidence intervals overlap the incumbent or evaluation cost exceeds the approved ceiling.
Decision rule
Reference only until a fresh target justifies its evaluation cost.
Learning exercise
Simulate one two-parameter ES update on a deterministic toy reward.
Explain the evaluation-cost tradeoff versus backpropagation.
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.