Reader Format Pool / V1
A sixteen-format open benchmark gave V72 97.41% complete reading, or 98.81% with the rewrite. Across the thirty-two independent formats encountered at that point, unrewritten mean complete reading was about 89.43%. The pool’s reserved half was subsequently opened by V77.
CONTROL COMPARISON
Create thirty-two new teacher-authored formats, mechanically validate them, and split into sixteen open and sixteen reserved formats. Score the open half on 400 already-executed configurations, producing 6,400 texts. This is measurement only; no reader is trained.
No qualification gate applies. The open pool provides a broader distribution than one test set but remains development data. The rewrite improves the open benchmark without becoming qualified. The original reserved half was unopened at this report and was later consumed by V77; it is no longer fresh qualification data.
- V72 · open half
- 97.41% fully right; worst format 77.50%
- V72 with rewrite · open half
- 98.81% fully right; worst format 91.00%
- Earlier second / third sets
- 91.52% / 71.39% fully right
- All thirty-two formats encountered then
- Approximately 89.43% fully right without rewrite
Question and method
Create thirty-two new teacher-authored formats, mechanically validate them, and split into sixteen open and sixteen reserved formats. Score the open half on 400 already-executed configurations, producing 6,400 texts. This is measurement only; no reader is trained.
Recorded results
| Condition | Recorded result |
|---|---|
| V72 · open half | 97.41% fully right; worst format 77.50% |
| V72 with rewrite · open half | 98.81% fully right; worst format 91.00% |
| Earlier second / third sets | 91.52% / 71.39% fully right |
| All thirty-two formats encountered then | Approximately 89.43% fully right without rewrite |
Comparative standing and locked gates
No qualification gate applies. The open pool provides a broader distribution than one test set but remains development data. The rewrite improves the open benchmark without becoming qualified. The original reserved half was unopened at this report and was later consumed by V77; it is no longer fresh qualification data.
Interpretation and limitations
Teacher templates and one fixed configuration family, not unrestricted real text. The aggregate combines differently sized sets and should be read with per-set and worst-format results. V77 documents later reserved-half failure.
SOURCE PROVENANCE
Reader format pool v1: the V72 reader on a wide base of formats
LABORATORY REPORT / 2026-10-08SOURCE CHECKSUM / SHA-256
9201ebc65d0d621661f96a578cb48460b561eb36edd3183df0fafa892d039ebbPublic journal edition reviewed 2026-10-09. Source documents and saved evidence were inspected; experiments were not rerun for this edition. Proprietary implementation code, model binaries, private infrastructure, and detailed machine records are not published here. Journal identifiers are editorial references. Catalog inclusion does not imply qualification or runtime promotion.