Target related hard-negative contexts
This commit is contained in:
@@ -19,6 +19,11 @@
|
||||
sampling assigned 75.5% of entries to Flanders. The cap retains every unique
|
||||
tile, removes repeats only and writes pre/post regional counts into the
|
||||
checksummed sampling evidence.
|
||||
- Added semantic hard-negative families after the v31 calibration isolated
|
||||
Oostende coastal/port precision as the dominant Flemish error. Coastal,
|
||||
industrial, ribbon, rural and regional-architecture failures now target
|
||||
related train-only negative contexts, with diagnostic repeats preserved
|
||||
ahead of generic repeats when the regional cap applies.
|
||||
|
||||
## 2026-07-27 - Guest demo and product professionalization
|
||||
|
||||
|
||||
Reference in New Issue
Block a user