GEPA adapts a model with prompts and labeled examples
A publication dated September 19, 2026 describes a GEPA demonstration adapting a commercially available model to specific decision criteria by changing its prompt and labeling examples.
The September 19, 2026 publication describes a demonstration using GEPA to adapt the commercially available Jev model, trained on synthetic data, to specific decision criteria. The presented approach changes the prompt and labels some examples; the text provides no metrics or quantitative results.
The material may interest readers studying ways to guide model decisions, but it does not show that the approach suits every task. Consult the original publication to check the demonstration details, and validate results with representative examples and defined criteria before applying the idea. If you use AI to study internal material, follow your organization’s policy: Rota Nacional detects personal data before execution and applies the configured policy; usage metadata remains in the dashboard, while raw prompts and responses stay in the AES-256-GCM-protected audit archive.