Task — engineering-spec@1

"Calibrated abstention - confidence output + not-in-memory behavior + false-memory metric"

draftTASK-MEMORY-254

module memory · class product · priority p1 · created 2026-07-08 · shipped null
depends on TASK-MEMORY-245 · blocks none

TASK-MEMORY-254: Calibrated abstention - confidence output + not-in-memory behavior + false-memory metric

1. Description

recall exposes calibrated confidence; benchmark tracks abstention bucket; false-memory rate on the ops tile

Migrated 2026-07-08 from the memory improvement backlog, folded into the task system as class: improvement. Source report refs: R57.

Acceptance criteria

[ ] recall exposes calibrated confidence; benchmark tracks abstention bucket; false-memory rate on the ops tile