README.md

metadata

pretty_name: MoralityLab Public Research Snapshot
license: mit
task_categories:
  - text-generation
  - text-classification
language:
  - en
tags:
  - moralitylab
  - alignment
  - constitutional-ai
  - storyworld
  - trm

MoralityLab Public Research Snapshot

Dataset Summary

Public-safe snapshot of MoralityLab research artifacts used for reproducible reporting:

• run manifests,
• adapter/TRM indexes,
• selected papers and docs,
• dashboard-facing summary JSON.

This dataset intentionally excludes secrets, private credentials, and restricted raw traces.

Intended Uses

• Public grant/research context.
• Dashboard demo payloads for Harness/Gym.
• Lightweight reproducibility receipts.

Out of Scope

• Sensitive logs.
• Full private training corpora.
• Confidential project metadata.

Data Fields

• source_path: original local path (for provenance only).
• copied_at: UTC time copied into this dataset.
• artifact_kind: paper | manifest | index | summary.
• run_id: optional run identifier.
• module: optional eval module.
• notes: optional curation notes.

Curation Process

1. Export from local research roots.
2. Redact and filter to public-safe fields.
3. Package into dataset snapshots and index CSV/JSONL.
4. Publish via Hugging Face Hub.

Limitations

• Snapshot-based: not a complete historical ledger.
• Some metrics are aggregated or inferred from manifests.
• Paths reflect local source topology and may not be directly accessible.

Citation

If you use this dataset, cite MoralityLab and link this dataset repository.