---

name: continual_recovery_evaluator description: Evaluate mechanism-faithful continual-learning recovery metrics and gate proxy evidence.

Continual Recovery Evaluator

When To Use

Use this skill when reconstructing the episodic-memory lifelong language learning method from de Masson d'Autume et al. without relying on an original implementation repository.

Inputs

• Small text examples represented as dictionaries.
• Deterministic configuration values such as memory probability, replay interval, K, and adaptation steps.

Outputs

• Auditable Python objects or JSON-serializable dictionaries matching the module contract.
• Logs that expose mechanism execution rather than only final metrics.

Workflow

1. Compute baseline and episodic retained accuracy on the same labels.
2. Report gain over baseline as the proxy metric.
3. Require explicit mechanism booleans before accepting proxy success.

Validation

Run python tests/test_continual_recovery_evaluator.py from this skill directory.

Limitations

The scripts are deterministic proxy utilities for bounded recovery. They do not train BERT or reproduce the full paper-scale datasets.