calculation trainer
Freeze authority, policy, and provenance
Bind a curated problem, solution graph, marking policy, and review state before making correctness or ECF claims.
1
Read with a clear lens
- versioned problems and policies
- source evidence
- review status
- fail-closed unsupported claims
Pause and write
What exact evidence would be required before calling one marking outcome correct?
2
Open the canonical material
After reading
Create a provenance chain from a learner step to the problem, policy version, source, and review status.
Knowledge check
When must a marking or ECF decision fail closed?
Write an answer to unlock the reference
It must return review required whenever the frozen problem and marking policy do not explicitly cover the observed path, or when required source or review provenance is missing.
AI_DRAFT · Policy behavior is an engineering contract pending qualified review.Study information and required outputs
Item information
- Stable ID
sprint.authority- Version
- 1.0.0
- Content status
- ready
- Kind
- reading
Dependencies
Model solution graphs and student stepsRequired repository outputs
docs/content/calculation_problems_seed.jsonFrozen problem, solution graph, and policy fixture
Done criteria
- Every claim names problem and policy versions
- Review state is visible
- Unsupported policy paths fail closed
Finish the session