Metraipolis › archive › dc_4faef6e866dd

Subliminal Prompting Beyond Static Geometry: Causal Depth and Multi-Token Confounds — arXiv:2609.19149

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.19149
License (as archived): CC0-1.0
Archived: 2026-10-10T03:45:50.152Z
SHA-256 of the archived text: 5562ab093e2f00fa85ccd15eaa52fa9a54a0f560932bc393eba19cb065eee471
Stable URL of this page: https://metraipolis.com/documento/dc_4faef6e866dd

How to cite

Subliminal Prompting Beyond Static Geometry: Causal Depth and Multi-Token Confounds — arXiv:2609.19149 — arXiv (en), https://arxiv.org/abs/2609.19149 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4faef6e866dd (Metraipolis archive dc_4faef6e866dd, sha256 5562ab093e2f00fa85ccd15eaa52fa9a54a0f560932bc393eba19cb065eee471)

BibTeX

@misc{dc4faef6e866dd,
  title = {Subliminal Prompting Beyond Static Geometry: Causal Depth and Multi-Token Confounds — arXiv:2609.19149},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.19149},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4faef6e866dd; sha256 5562ab093e2f00fa85ccd15eaa52fa9a54a0f560932bc393eba19cb065eee471},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.19149. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Subliminal learning shows that language models can transmit a hidden trait through outputs that appear unrelated to it. One proposed explanation, token entanglement, links animal and number tokens through the model's output vocabulary. Yet existing measurements answer different questions: whether outputs co-vary, fixed output vectors align, an answer can be read from a hidden state, or that state causally controls the answer. We measure each separately in a fixed animal-number prompting protocol. From Llama-3.1-8B to 70B, fixed output-vector similarity predicts behavior less well: the paired mean correlation change is -0.080 (95% CI [-0.127, -0.035]). A fixed output-head readout shows no resolved change in normalized depth AUC. To test control, we copy the temporary answer-position state from one number prompt into another at five depths and measure which prompt the final animal score follows. Donor-control AUC rises from 0.254 to 0.540, a paired change of +0.286 (95% CI [+0.272, +0.300]), with increases for all 18 concepts. The contrast remains with exactly eight transformer blocks remaining, while specificity and identity controls remain small or exact. In two Qwen models, scoring every digit in sequence does not recover the positive one-token association. Per-token averaging instead creates a positive pooled association that disappears after controlling number width, revealing a length confound. Thus, fixed geometry, observational readability, causal timing, and multi-token measurement are distinct properties of this frozen prompting channel. They constrain token-level explanations but do not identify the mechanism of training-time trait transfer.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-18 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.19149 — 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 2026-10-10T03:24:08.101Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).