Metraipolis › archive › dc_95f94bc0ec77
When Answer Decodability Is Not a Decision: Separating Final-Answer and Imminent-Emission Signals in Chain-of-Thought — arXiv:2611.11892
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.11892
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-10T18:45:08.481Z
- SHA-256 of the archived text
82f686ff767bf77537c6d48082a0e861040c37b1104f90cf0c7898f8581571ff- Stable URL of this page
https://metraipolis.com/documento/dc_95f94bc0ec77
How to cite
When Answer Decodability Is Not a Decision: Separating Final-Answer and Imminent-Emission Signals in Chain-of-Thought — arXiv:2611.11892 — arXiv (en), https://arxiv.org/abs/2611.11892 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_95f94bc0ec77 (Metraipolis archive dc_95f94bc0ec77, sha256 82f686ff767bf77537c6d48082a0e861040c37b1104f90cf0c7898f8581571ff)
BibTeX
@misc{dc95f94bc0ec77,
title = {When Answer Decodability Is Not a Decision: Separating Final-Answer and Imminent-Emission Signals in Chain-of-Thought — arXiv:2611.11892},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.11892},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_95f94bc0ec77; sha256 82f686ff767bf77537c6d48082a0e861040c37b1104f90cf0c7898f8581571ff},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.11892. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Answer probes applied to chain-of-thought hidden states are often read as evidence that a language model has already decided its final response. This reading overlooks a competing signal, the candidate the model is about to mention. We introduce emission-controlled dual-target probing (ECDP), which fits the same low-rank reader family separately to the imminent mention and to the final answer at identical positions, with target-specific nulls and thresholds fixed before any fit. On correct DecisionDepth (DD) rollouts in which a distractor is mentioned before the final answer, Qwen3.5-4B decodes the next mention at .966 (null .234) but the final answer only at .221 (null .186), an emission gap of .696 [.648, .744]; Llama-3.1-8B gives .575 [.514, .632]. This pattern is not universal. In a pre-specified evaluation of Qwen3-8B with thinking enabled on held-out MMLU questions, using a full-prefix attention reader over every thinking position, both targets are decodable and the final answer is read far more strongly (mention-minus-answer lift -.398 [-.459, -.336] at layer 22 over 467 questions). Model, task, positions, and reader change together between these settings, and an attempted matched comparison on DD produced no valid statistics because the imminent-mention probe did not train reliably, so we do not identify what reverses the order. On DD, planted-signal controls show that the reader recovers an injected final-answer signal at one tenth of full dose. Answer-candidate decodability is therefore target-relative and protocol-dependent: a strong answer readout does not by itself establish when a model commits, and a probe that fails to train is not evidence that the information is absent. ECDP is a practical control for distinguishing imminent text emission from evidence of a settled decision. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno). Categorie arXiv: Machine Learning (cs.LG) ; Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2611.11892 — 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 2026-11-10T18:37:52.480Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).