Source-aware memory and revision checks

Question

Can a retrieved memory be traced to its source and distinguished from an obsolete revision? SeraLab investigates this locally with explicit source identity and revision handling.

Reproducible evaluation procedure

1. Create synthetic documents with known source identifiers and one deliberately repeated document.
2. Import, interrupt and resume the process; count persisted entries and duplicates.
3. Revise one document, then mark a previous item withdrawn. Query both old and new wording.
4. Inspect returned source/version links and verify that withdrawn content is not silently presented as current.
5. Record retrieval hits, wrong-version hits, duplicate counts, recovery behavior and elapsed time separately.

Environment and evidence limits

Implementation checks used a local Windows environment, synthetic records and local hybrid retrieval. This note publishes the procedure, not the private records, model files or raw logs. No independently reproduced benchmark, hardware-normalized throughput, recall score or latency claim is available. A real personal archive has not been treated as validated merely because synthetic checks pass.

Next evidence

A small shareable synthetic corpus with labeled source/version expectations, pinned environment details and repeated measurements is needed before comparative quality claims. Sensitive data should remain outside the public website and source repository.