| name | student-analyze-results |
| description | Convert metrics into supported claims, anomalies, and next-step decisions. |
Student Analyze Results
Purpose
Turn raw metrics into honest claims and a credible next experiment.
Workflow
- Read the experiment plan and the result artifact together.
- Compare against the intended baselines and ablations.
- Separate fair comparisons from noisy or compromised ones.
- Identify stable gains, fragile gains, and outright failures.
- Publish to your canonical shared artifact directory:
{artifact_dir}/{your-name}/analysis_{experiment_id}.md
Output Format
# Result Analysis: {title}
## Inputs
- **Plan**: ...
- **Results**: ...
## Headline Findings
1. ...
## Comparison Table
| Comparison | Fair? | Outcome | Notes |
|---|---|---|---|
## Supported Claims
1. ...
## Unsupported but tempting claims
1. ...
## Anomalies and Failure Modes
1. ...
## Recommended Next Experiment
1. ...
Rules
- Every claim must point back to a concrete result artifact.
- If the baseline comparison is unfair or noisy, say so explicitly.
- Always include one section called
Unsupported but tempting claims.