up
This commit is contained in:
11
bench/reachability-benchmark/tools/scorer/README.md
Normal file
11
bench/reachability-benchmark/tools/scorer/README.md
Normal file
@@ -0,0 +1,11 @@
|
||||
# rb-score (placeholder)
|
||||
|
||||
Planned CLI to score reachability submissions against truth sets.
|
||||
|
||||
Future work (BENCH-SCORER-513-008):
|
||||
- Validate submission against `schemas/submission.schema.json`.
|
||||
- Validate truth against `schemas/truth.schema.json`.
|
||||
- Compute precision/recall/F1, explainability score (0-3), runtime stats, determinism rate.
|
||||
- Emit JSON report with stable ordering.
|
||||
|
||||
For now this folder is a stub; implementation will be added in task 513-008 once schemas stabilize.
|
||||
Reference in New Issue
Block a user