---
id: workflow.weights-training.learning-targets
title: Define learning targets from policies and rubrics
summary: Turn an observed behavioral distinction into a specific teaching purpose.
kind: task
product_area: weights_training
status: stable
updated: 2026-11-11
canonical: /docs/weights-training/learning-targets
---

# Define learning targets from policies and rubrics

Turn an observed behavioral distinction into a specific teaching purpose.

## Prerequisites

Reviewed criteria, their applicability, and representative failures or missing behavior.

## Steps

1. Identify the requirement the learner should satisfy and the condition that makes it applicable.
2. Inspect failures and successful neighboring Cases so the objective includes exceptions and protected behavior.
3. Select the intended benchmark criteria and scope in the training preparation.
4. Inspect whether generated examples teach the distinction through their actual targets. Keep protected assessment Cases separate from learner material.

## Object and state changes

Selecting criteria guides the preparation. It does not alter the governed standard or establish that a trained model learned it.

## Success criteria

The team can explain what each example teaches and which independent assessment can test the resulting behavior.

## Common failure modes

Using a broad score as the learning target; copying protected benchmark Cases into the learner dataset.

## Related reference pages

{% related-card-grid title="Continue this workflow" %}
- [Prepare data for weights training](/docs/weights-training)
{% /related-card-grid %}

## Related troubleshooting pages

{% related-card-grid title="Resolve a problem" %}
- [Synthetic Cases That Feel Unrealistic](/docs/troubleshooting/unrealistic-synthetic-cases)
- [Unbalanced Coverage](/docs/troubleshooting/unbalanced-coverage)
{% /related-card-grid %}