Erratum28 August 2026

The right lift, the wrong denominator

QueryPilot's correction result belonged to the core set, not the combined benchmark.

What happened

The correction loop moved the core result from 63 successful executions to 67. The denominator for that comparison is 70 core queries.

I had previously placed the lift beside the total that also included 12 adversarial queries. The percentage was arithmetically correct for the core run and contextually wrong where I used it. The two sets now stay separate.

Attached measurements

The denominator stays with the number.

63 to 67 of 70: QueryPilot's correction loop moved execution success from 63 to 67 queries on the 70-query core set.; 70 core benchmark queries

Source boundary

What this record was checked against

01artifact
Reviewed internally. The repository or artifact is private, so its path is not published.