---
id: workflow.weights-training.synthesize-examples
title: Synthesize training examples with Lemon Weave
summary: Expand reviewed expert standards into examples with a concrete teaching purpose.
kind: task
product_area: weights_training
status: stable
updated: 2026-11-11
canonical: /docs/weights-training/synthesize-examples
---

# Synthesize training examples with Lemon Weave

Expand reviewed expert standards into examples with a concrete teaching purpose.

## Prerequisites

A preparation with the intended Benchmark Version, method, selected scope, and generation settings.

## Steps

1. Choose the requirements and conditions that need additional demonstrations or preferences.
2. Generate through Weights training and inspect the actual examples in Dataset.
3. Check the input, target response or preference pair, source context, and governing criteria. Exclude examples that fail to teach the intended distinction.
4. Inspect the recorded dataset state and Coverage before preparing a package.

## Object and state changes

Generated examples remain part of their training preparation. Generation completion, dataset review, provider training, and held-out assessment are separate outcomes.

## Success criteria

The prepared material expresses the intended expert distinction without depending on leaked assessment answers or unsupported context.

## Common failure modes

Accepting examples from status alone; confusing expected labels with observed judgments; silently dropping material a provider cannot represent.

## Related reference pages

{% related-card-grid title="Continue this workflow" %}
- [Inspect dataset quality and coverage](/docs/weights-training/datasets)
{% /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 %}