ablation_studies

Scope: Ablation study matrix.
Audience: Reviewers and contributors.
Guarantees: Ablations are explicit.
Non-Goals: Full experiment suite.
Why: This doc exists to record its single responsibility for review.

Overview

This doc lists required ablations.
Architecture context lives in Architecture.
Claim context lives in Falsifiable Claim.

Contracts

Each ablation removes one capability.
Ablations compare recovery and failure metrics.
Metrics follow Metrics.
Evidence uses tests/regression/test_negative_results.py.
| Ablation | Removed capability | Expected signal. |
| --- | --- | --- |
| No regulator | Proposal application | Recovery drop. |
| No stochasticity | Noise in transitions | Variance drop. |
| No recovery | Failure handling | Recovery drop. |
| No constraints | Constraint checks | Failure rise. |

Invariants

Ablation list stays fixed.
List aligns with Core.
Evidence aligns with tests/regression/test_negative_results.py.

Failure Modes

Missing ablations weaken claims.
List drift breaks Core.
Missing evidence breaks Docs Style.

Extension Points

Ablation updates follow Docs Style.
Extensions align with Experimental.
Evidence updates align with tests/unit/test_docs_contract.py.

Exit Criteria

This doc is obsolete when ablations are encoded.
The replacement is Metrics.
Obsolete docs are removed.

Code refs: tests/regression/test_negative_results.py.