Finding

Sign in with GitHub
← Publications

Finding · P44 · Author-curated

The vocabulary sizes of all nine tokenizers at the pinned revisions equal those printed in Table 1 of arXiv:2604.10799v1.

Published by @stw2 · 2026-09-14 · Sources, measurements and interpretation are supplied by the author.

Structured assertion

Relation: Matches printed values

quantity
Vocabulary sizeQuantity checked.
tokenizers matching
9 tokenizersTokenizers whose value equals the printed value.
tokenizers total
9 tokenizersTokenizers checked.

Experimental provenance

Method and evaluation protocol
Load each tokenizer at its pinned revision and compare its vocabulary size with Table 1.
Dataset
Tokenizer files of nine Hugging Face repositories at pinned revisions.Version: unspecified · Access: restricted
Reported results
9 of 9 vocabulary sizes equal the printed values.
Uncertainty and replication
Exact counts.
Limitations
Two of the repositories are gated.

Paper citations

arXiv:2604.10799v1 →Revision supplied by author
  1. Table 1, p. 3, Vocab Size column

Concept definitions

Reuse the defining version and key when the meaning fits your assertion.

Matches printed values

The measured quantity equals the published value for the stated number of items.

Key matches_printed · version 46157fd2-de49-49f3-9d8e-b91e77b915c2

Vocabulary size

Number of tokens in a tokenizer's vocabulary.

Key vocabulary_size · version 2d581473-7212-4ea2-bf80-f0c4b8cb247e

Exact references

supports

2d581473-7212-4ea2-bf80-f0c4b8cb247e

APT4 32,000 and the Mistral-derived tokenizer 32,128 are among the matching sizes.