Fine Tuning Experiment.task

# Task Prompt: Fine-Tuning Experiment

## Objective

Design and execute a reproducible fine-tuning experiment for {{goal_metric}} improvement.

## Required Inputs

- experiment brief
- dataset references
- base model profile
- allowed compute budget

## Execution Checklist

1. Define hypothesis and success metrics.
2. Run data quality and safety audit.
3. Execute training with versioned config.
4. Evaluate with baseline comparison.
5. Decide promote/hold using objective thresholds.

## Completion Criteria

- reproducible config and logs
- metric delta reported with confidence
- safety behavior not regressed