| --- |
| license: mit |
| language: |
| - en |
| pretty_name: EvalKit Tutorial v0.1 |
| size_categories: |
| - n<1K |
| tags: |
| - synthetic |
| - evaluation |
| - rubric |
| - tutorial |
| --- |
| |
| # EvalKit Tutorial v0.1 Dataset Card |
|
|
| This dataset is synthetic tutorial data for exercising local EvalKit workflows. It is not expert-authored, not human-validated, not benchmark-grade, and not a safety or model-quality certification. |
|
|
| ## Intended Use |
|
|
| Use these small fixtures to run `evalkit validate-rubric`, `evalkit lint-rubric`, `evalkit score`, report generation, and adapter mapping examples without an AuraOne account or API key. |
|
|
| ## Out Of Scope Use |
|
|
| Do not use this data to compare models, publish benchmark claims, certify safety, or represent customer reviewer performance. |
|
|
| ## Files |
|
|
| - `rubric.jsonl`: synthetic code-review tutorial rubric. |
| - `model_outputs.jsonl`: synthetic tutorial model outputs. |
| - `labels.jsonl`: synthetic tutorial labels. |
| - `expected_scores.json`: deterministic expected score summary. |
| - `LICENSE`: MIT-compatible fixture license. |
|
|
| ## Synthetic Generation Process |
|
|
| Rows were hand-written as low-risk code-review-style tutorial examples. They intentionally avoid medical, legal, safety-critical, customer, reviewer identity, and production data. |
|
|
| ## Limitations |
|
|
| - Small sample count. |
| - Not validated by domain experts. |
| - Public tutorial data can leak into prompts and should not be used as a hidden eval. |
|
|
| ## Citation |
|
|
| Cite AuraOne EvalKit documentation when reusing the template or adapting the schema. |
|
|
