Significance Levels (+2)
Significance Levels
| Level | Meaning | Action |
|---|---|---|
match |
Values identical (within tolerance) | None |
cosmetic |
< 0.01% numeric diff | Safe to ignore |
minor |
0.01-1% numeric diff | Review if important |
significant |
> 1% numeric diff | Must investigate |
missing |
Present in mono, absent in mod | Fix extractor/builder |
extra |
Absent in mono, present in mod | Fix builder defaults |
type_mismatch |
Different types (bool vs string) | Fix type handling |
Running Semantic Validation
# Single model
uv run python scripts/semantic_validate.py model.yml modular_dir/
# With HTML report
uv run python scripts/semantic_validate.py model.yml modular_dir/ --html report.html
# Batch mode
uv run python scripts/semantic_validate.py model.yml modular_dir/ --batch output_dir/
Integrated in Benchmark
The benchmark pipeline (scripts/benchmark_model_library.py) runs semantic validation as a pre-statics gate:
.dat → YAML → extract → spec → generate modular → [SEMANTIC CHECK] → [statics] → compare