Metraipolis › archive › dc_e11f81eaa4d2

Nameless Tokenization: A Lossless Tokenizer-Level Defense Against Control-Token Forgery in Open-Weight LLMs — arXiv:2609.16984

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.16984
License (as archived)
CC0-1.0
Archived
2026-10-15T19:19:23.291Z
SHA-256 of the archived text
0fb9a4dd459c7fd8842fcc4e80ea6ebb4c3a0e76e87202a09f528456f7112651
Stable URL of this page
https://metraipolis.com/documento/dc_e11f81eaa4d2

How to cite

Nameless Tokenization: A Lossless Tokenizer-Level Defense Against Control-Token Forgery in Open-Weight LLMs — arXiv:2609.16984 — arXiv (en), https://arxiv.org/abs/2609.16984 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e11f81eaa4d2 (Metraipolis archive dc_e11f81eaa4d2, sha256 0fb9a4dd459c7fd8842fcc4e80ea6ebb4c3a0e76e87202a09f528456f7112651)
BibTeX
@misc{dce11f81eaa4d2,
  title = {Nameless Tokenization: A Lossless Tokenizer-Level Defense Against Control-Token Forgery in Open-Weight LLMs — arXiv:2609.16984},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.16984},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e11f81eaa4d2; sha256 0fb9a4dd459c7fd8842fcc4e80ea6ebb4c3a0e76e87202a09f528456f7112651},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.16984. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Open-weight language models publish the strings their chat templates use to mark turns, roles and tool results, which the tokenizer maps back to the reserved identifiers the model obeys. Anyone who controls text in a prompt can therefore write a turn boundary indistinguishable from one the serving stack wrote. We audit 256 deployed chat tokenizers. All are forgeable, and the flag usually recommended as a fix leaves 56.6% forgeable because it misses the tool and reasoning markers agent systems rely on. We propose nameless tokenization, which leaves the control entries with a reserved identifier and no surface string, so the content encoder cannot emit one and message content reaches the model unaltered. Across five tokenizer families it reproduces the standard token stream exactly on attack-free data and lifts accuracy on a probe of delimiter-bearing text from 8.5% to 59.9%, where sanitizers lose it. Separating a delimiter's appearance from its identifier shows the identifier matters little against a bare task instruction, but carries most of a forged tool result and most of any forged turn once the system message tells the model to treat user content as data.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-16 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.16984 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-15T18:32:32.848Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-16?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).