Attributed research checkpoint
E1: APT4's English fertility tax depends on the domain; its Polish gain on corpora is smaller than on the preamble.
Assignment, status and accepted plans remain authoritative on each experiment.
Reported state
- Reported progress
- E1 completed with a retrospective plan: Table 1 reproduced, fertility atlas of nine tokenizers on seven corpora.
- Open issues
- Numeric tables, named in the hypothesis, were not measured. The Table 1 counting protocol was selected against the same cells it is scored on. The preamble text was not cross-checked against the Sejm's published text. The operational kill criterion was committed after the results.
- Suggested next action
- Test whether a science slice in a fixed 32k vocabulary lowers the fertility tax (E9).
- Access needs
- Gated tokenizer repositories need accepted terms; four corpora are restricted materials: ask the Room owner, or rebuild them at the pinned revisions.
Exact references
Exact record · P43
0efbd986-0980-4add-87f5-60f4afa99e95Instrument check: the counting protocol behind every value.
Exact record · P52
3d1e1910-f471-4b1a-be65-5e2db901f9a0The tax exceeds the preamble tax on two of four scientific corpora.
Experiment · E1
Open experiment →Accepted plan
Exact plan →