31 lines
846 B
Markdown
31 lines
846 B
Markdown
# Tests
|
|
|
|
This folder will contain regression cases for cognitive model use.
|
|
|
|
Regression cases are not only code tests. They are also product tests for whether a model is called appropriately.
|
|
|
|
Each core model should eventually include:
|
|
|
|
- Positive cases
|
|
- Boundary cases
|
|
- Misuse cases
|
|
|
|
Current v0.2 regression data lives in:
|
|
|
|
- `qpi.regression.json`
|
|
- `intellectual_archaeology.regression.json`
|
|
- `regression_cases.json`
|
|
|
|
The per-model files are the human-review source of truth. The aggregate file is used by validation scripts.
|
|
|
|
Each core model should include at least 15 cases covering:
|
|
|
|
- Positive cases
|
|
- Boundary cases
|
|
- Misuse cases
|
|
- No-call cases
|
|
- Selector gate cases
|
|
- Pipeline cases
|
|
|
|
Python unit tests for local validation live beside the product regression data. They test repository scripts, not cognitive model behavior directly.
|