Feedback State Matrix v1
Deterministic evaluator-state routing for the recommendation contract.
The Single Idea
Evaluator and replay states were routed through the feedback recommendation contract so pass, revision, model-review, evaluator-pending, and start-path states each have an auditable next action.
01
Snapshot
5cases
5valid
5re-ranked
1model review
1advance
1revise
States covered: awaiting_evaluation, decision_generated, evaluated_pass, evaluated_review, needs_revision.
02
Matrix
| State | Action | Strategy | Module | Valid | Failures |
|---|---|---|---|---|---|
| decision_generated | await_practice_submission | decision_generated_original_path | ed46a9a7b5d5e472 | yes | none |
| awaiting_evaluation | run_evaluator | awaiting_evaluation_original_path | ed46a9a7b5d5e472 | yes | none |
| needs_revision | revise_current_module | failed_gate_remediation | ed46a9a7b5d5e472 | yes | none |
| evaluated_review | model_review_before_advancing | model_review_hold_current | ed46a9a7b5d5e472 | yes | none |
| evaluated_pass | advance_after_model_review | evaluation_pass_remaining_path | c0c67449115f91b8 | yes | none |
03
Boundary
This is deterministic local evaluator-state proof. It does not claim adaptive learning, mutate planner output, or use production telemetry.
Where to start
Use this matrix to keep pass, revision, review, pending-evaluation, and start-path behavior explicit as the learner loop matures.
Use this matrix to keep pass, revision, review, pending-evaluation, and start-path behavior explicit as the learner loop matures.