EXECUTED · V4
Comparing knowledge-revision methods
When one piece of evidence is withdrawn, which systems retract the affected conclusions without discarding unaffected information?
Original verbatim terminal output has not been independently established; the structured result below is not original console output.
| Method | Linked conclusions invalidated | False retentions |
|---|---|---|
| Weak document memory | 0/12 | 14* |
| Conventional dependency tracker | 12/12 | 0 |
| Experimental CRSE kernel | 12/12 | 0 |
*Fourteen refers to the broader ground-truth comparison, not just the twelve linked dependents. Both strong methods also had zero false invalidations and preserved unaffected conclusions. This is a structured rendering of the archived result, NOT a verbatim terminal transcript.
Structured scientific record and disclosure status
Verified three-arm bounded synthetic comparison. CRSE and conventional dependency tracking each invalidated 12/12 linked conclusions. No superiority over the conventional graph baseline was established.
| Method | Linked conclusions invalidated | False retentions |
|---|---|---|
| Weak document memory | 0/12 | 14* |
| Conventional dependency tracker | 12/12 | 0 |
| Experimental CRSE kernel | 12/12 | 0 |
Original verbatim terminal output has not been independently established; the structured result below is not original console output.
Explanatory replay of the recorded comparison
Teaching representation, not a disclosed internal graph or live execution.
Method, evidence provenance and scientific limitations
One synthetic challenge with twelve explicitly defined dependents, compared across intentionally weak document memory, conventional graph tracking and a CRSE kernel. The original full verbatim terminal transcript and exact scientific execution time have not been independently verified. This was not an automatic dependency-discovery task, field test, calibrated performance study or independent-laboratory replication. Graph-parity is the finding, not a competitive advantage.
Execution date: exact scientific execution timestamp NOT_VERIFIED; archive-context dates do not establish execution time. Public publication date: NOT PUBLISHED.
Release note: First three-arm bounded comparison; conventional graph and CRSE both matched the specified ground truth.