# References and provenance

## Primary structure reference

- Blackfrost AI, **DWM (Domain Weight Modification) by Blackfrost_AI — Qwen 9B
  runbook package**:
  <https://github.com/Blackfrost-AI/blackfrost-dwm-qwen9b-runbook>

The public runbook supplied the structural model for this course: start-here
orientation, method overview, pinned identity, staged runbook, fail-stop gates,
and a reporting template. This learning pack is a separate educational work. It
does not reproduce the runbook’s raw paired bank or executable scripts.

## Model reference

- Qwen, **Qwen3.5-9B model repository**:
  <https://huggingface.co/Qwen/Qwen3.5-9B>

The historical course case pins the exact revision recorded in
`CASE-FILE.json`.

## Concepts to review

- Transformer residual streams and linear projections
- Rank-1 matrix updates
- Frobenius norms
- Cryptographic file digests
- Reproducible experiment manifests
- Independent artifact verification
- Behavioral evaluation and release gates

## Evidence note

The case file is a sanitized educational record. The absence of raw weights,
paired content, activations, and executable editing tools is deliberate. The
record is suitable for practicing audit reasoning, not for recreating an
intervention or claiming deployment qualification.