45 lines
1.4 KiB
Markdown
45 lines
1.4 KiB
Markdown
# Extraction Quality Audit
|
|
|
|
## 1. Status
|
|
|
|
```text
|
|
status: passed
|
|
audit_type: material extraction quality
|
|
material_units_total: 29
|
|
topic_docs_with_extraction: 5
|
|
root_doc_updated: pending_if_downstream_requires
|
|
```
|
|
|
|
## 2. Checks
|
|
|
|
| check | result | note |
|
|
| --- | --- | --- |
|
|
| Step 0 confirmed before extraction | passed | `human-confirmation.md` recorded |
|
|
| Material units preserve source block IDs | passed | each unit lists `source_blocks` |
|
|
| No flat topic regression | passed | extraction follows A/B/C/D/E graph |
|
|
| QPI relation preserved | passed | QPI kept as GL3 problem-representation tool |
|
|
| Giant Ark relation preserved | passed | formal model replacing Giant Cognition 2.0 |
|
|
| Reality Lab relation preserved | passed | action module inside intention-action-feedback |
|
|
| Family Triangle relation preserved | passed | independent model and scenario model called by Giant Ark |
|
|
| Adolescent weighting caveat preserved | passed | simplifying assumption only |
|
|
|
|
## 3. Coverage By Layer
|
|
|
|
```text
|
|
A: extracted
|
|
B: extracted
|
|
C: extracted
|
|
D: extracted
|
|
E: extracted
|
|
ROOT: overview only; downstream tracks should read _material-extraction-index.md
|
|
```
|
|
|
|
## 4. Residual Risk
|
|
|
|
```text
|
|
residual_risk: medium
|
|
reason: extraction is paraphrased material-unit level, not quote-level evidence extraction
|
|
```
|
|
|
|
If the next task is publication, formal model-card writing, or audit-grade evidence review, create a quote-level evidence pack from the original source before final claims.
|