# Start here

## Course

**DWM Foundations: Method, Evidence, and the Qwen 9B Case**

- Level: Intermediate
- Estimated time: two hours
- Format: self-paced or instructor-led
- Prerequisites: transformer fundamentals, basic linear algebra, and comfort
  reading structured technical evidence

## Learning outcomes

By the end, you should be able to:

1. Distinguish Domain Weight Modification from fine-tuning, ablation, and
   prompt steering.
2. Define an immutable experiment identity using exact revisions, parameters,
   and cryptographic digests.
3. Explain paired prefill measurement, a bounded rank-1 projection, iterative
   recapture, and per-matrix norm restoration.
4. Audit fail-stop gates, exact controls, and independently loaded artifacts.
5. Calculate a relative metric reduction and describe only what it establishes.
6. Draft a short model-change record with explicit limitations and next tests.

## Recommended sequence

1. Read `COURSE-TRANSCRIPT.md` modules 1–3.
2. Complete workbook exercises 1–3.
3. Read transcript modules 4–5 and inspect `EVIDENCE-CARDS.md`.
4. Complete workbook exercises 4–5.
5. Read module 6 and inspect `CASE-FILE.json`.
6. Complete the capstone and knowledge check.
7. Compare your reasoning with `INSTRUCTOR-GUIDE.md`.
8. Draft a record with `MODEL-CARD-TEMPLATE.md`.

## Course boundary

This is an evidence-literacy course. The pack does not include model weights,
the raw measurement bank, activation captures, or executable editing tools.
Work only on artifacts you own or are explicitly authorized to evaluate.

The case file reports historical measurements. It does not establish broad
capability preservation, safety, or deployment readiness.