Metraipolis › archive › dc_ce5d3009ce2f
Decoupling Is Not Identification: Supervised Evidential Learning in Next-Token Prediction — arXiv:2609.26268
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.26268
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T13:45:43.254Z
- SHA-256 of the archived text
2e516bf223b19366ac5f3f408e69c6199ce151c16b2ee5902cb4faa3e20cc701- Stable URL of this page
https://metraipolis.com/documento/dc_ce5d3009ce2f
How to cite
Decoupling Is Not Identification: Supervised Evidential Learning in Next-Token Prediction — arXiv:2609.26268 — arXiv (en), https://arxiv.org/abs/2609.26268 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ce5d3009ce2f (Metraipolis archive dc_ce5d3009ce2f, sha256 2e516bf223b19366ac5f3f408e69c6199ce151c16b2ee5902cb4faa3e20cc701)
BibTeX
@misc{dcce5d3009ce2f,
title = {Decoupling Is Not Identification: Supervised Evidential Learning in Next-Token Prediction — arXiv:2609.26268},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.26268},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ce5d3009ce2f; sha256 2e516bf223b19366ac5f3f408e69c6199ce151c16b2ee5902cb4faa3e20cc701},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.26268. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
A next-token probability says what a model predicts, not how much training support lies behind it. A Dirichlet head can represent this distinction by separating mean $m$ from concentration $S$, but decoupling does not identify what $S$ means. Here we propose an Evidential Next-Token Prediction (ENTOP) framework to audit this gap on character-level Moby-Dick, using exact 8-gram count as a reproducible lexical-support label and withholding count regression from 20% of context types. Standard implicit evidential training carries essentially no count signal beyond confidence on held-out-label types (partial Spearman $\rho = 0.001 \pm 0.014$), whereas explicit supervision generalizes ($\rho = 0.201 \pm 0.010$; matched-pair win $= 0.822 \pm 0.021$). CE predictive entropy is at chance for unseen 8-grams (AUROC $= 0.490 \pm 0.004$), while supervised vacuity reaches $0.772 \pm 0.003$, comparable with an indexed CE-representation baseline ($0.769$) but below the tautological corpus oracle ($1.000$). Neither longest-suffix nor representation-distance strata explain where amortization succeeds. Increasing count weight under the digamma objective improves support fit only by sacrificing prediction. A constant predictor wins natural log-RMSE, and vacuity does not improve error deferral. These results motivate a minimum evidence protocol---confidence control, matched pairs, held-out labels, a constant baseline, and a decision test---and show that concentration can pass identification while failing calibration and utility. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-23 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.26268 — 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 2026-10-08T13:18:38.163Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-23?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).