From final scores to measurable corrections

Can saved Noetis records show whether a correction actually helped?

Audited existing Noetis evaluation exports and built an offline trace validator. The saved records reproduce, but the scoring endpoint cannot carry claims about architectural benefit.

• Replaying the current scorer reproduced all 119 nonempty stored scores.
• Appending the full source as a labeled appendix removed detected-issue credit in all 114 outputs that had it.
• Removing appended approved findings changed detection counts in 44 of 59 nonempty panel outputs.
• Saved results keep the final brief and latest check, but not before/after pairs, so a repair flag cannot distinguish a fix from a new error.
• Usage data existed for 19 panel cases and no baseline cases.

• E13-A Audit of saved Noetis evaluation records Executed (audit)
• E13-B Local revision pilot with evidence receipts Proposed

• H13-1 Evidence receipts at the revision boundary improve task success over ordinary reconsideration.

• Revision-drift guard. Flag synthesis that omits supported risks or adds untraced claims, rather than presenting it as consensus.

• Does evidence packaging change outcomes at a fixed state?
• Does Noetis outperform a strong single agent at matched evidence access and measured cost?

Builds on Round 12, Round 11. Continued in Round 20, Round 19, Round 18, Round 17, Round 16, Round 15, Round 14.