File size: 1,121 Bytes
29a12ae | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 | # Reproducibility
This Hugging Face package contains dataset and saved-output benchmark artifacts. The full code and runner scripts live in the GitHub repository:
https://github.com/SeanWan514/StableEval-Arena
No model API calls are required to inspect or validate this package.
## Local Package Validation
From the package root:
```bash
python scripts/validate_local_package.py
```
The validator checks required files, CSV/JSONL readability, case counts, prompt counts, prediction counts, the 507 label distribution, stale paths, cache files, and obvious secret patterns.
## Saved-Output Reproduction
Use the GitHub repository for saved-output metric recomputation scripts:
- `scripts/evaluate/recompute_stableeval_120_enriched.py`
- `scripts/evaluate/recompute_stableeval_507_natural.py`
Those scripts read the same canonical artifacts represented here and write recomputed outputs only to a user-specified local directory.
## Optional API Reruns
Agent reruns are not part of this Hugging Face package. They require OpenRouter credentials and the GitHub codebase. Do not write API keys into dataset files.
|