Metraipolis › archive › dc_84811db282db

Prompt-Model Interaction Reaches the Fixed Points: A deterministic, task-free structural readout -- and the factorizations of it that failed — arXiv:2608.21315

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.21315
License (as archived)
CC0-1.0
Archived
2026-10-18T10:18:34.997Z
SHA-256 of the archived text
8727bc37ee5dd939912054e930e6877e6fd513718c9db29ee9f1d0175243af01
Stable URL of this page
https://metraipolis.com/documento/dc_84811db282db

How to cite

Prompt-Model Interaction Reaches the Fixed Points: A deterministic, task-free structural readout -- and the factorizations of it that failed — arXiv:2608.21315 — arXiv (en), https://arxiv.org/abs/2608.21315 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_84811db282db (Metraipolis archive dc_84811db282db, sha256 8727bc37ee5dd939912054e930e6877e6fd513718c9db29ee9f1d0175243af01)
BibTeX
@misc{dc84811db282db,
  title = {Prompt-Model Interaction Reaches the Fixed Points: A deterministic, task-free structural readout -- and the factorizations of it that failed — arXiv:2608.21315},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.21315},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_84811db282db; sha256 8727bc37ee5dd939912054e930e6877e6fd513718c9db29ee9f1d0175243af01},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.21315. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

That a prompt's effect is not a property of the prompt is established: prompts optimised for one model degrade on another, and rankings reorder under neutral reformatting. That evidence is about task accuracy, which cannot say whether the interaction is a fact about task machinery or about the conditional distribution itself. We ask on a readout with no task in it: the fixed-point structure of the short-window argmax map x_{t+1} = argmax_x p(x | x_{t-1}, x_t), censused from 96 starts. It is deterministic, so nothing can be helped or hurt, and it exists only at short windows -- four of six models lose it entirely by window 16 -- so everything here concerns how a model reads a fragment. Two results. First, the interaction reaches this readout at full magnitude: nine tokens of conditioning move the fixed-point fraction across most of its range, change a four-way structural class, and reorder models, while instruction tuning worth 60.5 IFEval points moves the class by zero. Second, nothing we proposed carries it. Prefix length fails: the effect is not monotone. Four phenomenological factors -- prose-versus-markup, a universal direction, bidirectionality, instruct-resistance -- were each withdrawn within one run of being proposed, dissolved by widening the sample. And the nearest mechanistic account, attention-sink dominance of early tokens, predicts the sign of the shift on 2 of 5 models -- chance -- while a length-by-content cross shows it holds on real text and fails on our probe's uniformly random input, so we are outside its regime, not against it. One fixed nine-token prefix drives four models toward 0 and two toward 1; the bidirectionality survives in-distribution starts. On this readout the unit of explanation is the prompt-model pair. The recurring error it caught in us has a name: a criterion with a shape applied to a quantity with no room to vary.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-24 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2608.21315 — 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 2026-10-18T10:17:37.604Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-24?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).