At time zero, the 95% interval of the FVT−FOCUS gap of APT4 transplants of Qwen2.5-1.5B excludes zero on 7 of 9 domains and includes zero on math_clean statements and English arXiv abstracts.

Relation: Interval excludes zero on a count of domains

metric

FVT−FOCUS gapQuantity with an interval per domain.

model

APT4 transplant of Qwen2.5-1.5BModels measured.

setting

time zero: no training after embedding initialisation; tokenizers loaded through a whitespace canary with the pristine APT4 tokenizer as fallbackTraining state and scoring condition.

count excluding

7 domainsDomains whose interval excludes zero.

count total

9 domainsDomains with an interval.

domains including zero

math_clean statements; English arXiv abstractsDomains whose interval includes zero.

Method and evaluation protocol

Check whether each domain's 95% paired document bootstrap interval of FVT-minus-FOCUS bits per byte contains zero.

Dataset

First 200 documents of each of the nine primary evaluation domains: the Polish FineWeb2-HQ holdout, the English SlimPajama holdout, Polish reviews, Polish Wikipedia science articles, Polish PES examination questions, English arXiv abstracts, English LaTeX method sections, English Python code and math_clean statements.Version: unspecified · Access: restricted

Reported results

sci_latex: [0.0579, 0.0788]; sci_python: [0.0941, 0.1195]; math_clean: [-0.0003, 0.0277]; sci_arxiv: [-0.0028, 0.027]; pl_wiki_sci: [0.3906, 0.4174]; pl_pes: [0.2627, 0.3096]; pl: [0.3626, 0.4031]; en: [0.0514, 0.0793]; pl_informal: [0.3326, 0.3902]

Uncertainty and replication

Bootstrap of 1000 resamples, seed 20260705.

Evidence references

analysis.json · publichttps://github.com/stw2/tokenizer-science-tax/blob/e7d4bb2df26254f88ecbf390a7cbf69ac210e0e3/experiments/E07-embedding-init/results/analysis.jsonmetrics.json · publichttps://github.com/stw2/tokenizer-science-tax/blob/e7d4bb2df26254f88ecbf390a7cbf69ac210e0e3/experiments/E07-embedding-init/results/metrics.json

Limitations

The design text was first committed together with the stage-1 results. Init damage compares APT4 transplants with the base model under its own tokenizer at 2,048-token windows, which cover different byte spans, and fertility inflates absolute damage on English formal text. One build and one scoring run per initialisation at 1.5B.