Metraipolis › archive › dc_f4d793ba2f80
Labels Override Definitions in Jev-Style Typed Decision Models — arXiv:2610.02586
- Source
- arXiv
- Original
- https://arxiv.org/abs/2610.02586
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-05T22:49:04.333Z
- SHA-256 of the archived text
250cc8d225f47bfa72e8cb75a92be28573e9bb0875c4d8ce78eee8a8b756db50- Stable URL of this page
https://metraipolis.com/documento/dc_f4d793ba2f80
How to cite
Labels Override Definitions in Jev-Style Typed Decision Models — arXiv:2610.02586 — arXiv, https://arxiv.org/abs/2610.02586 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f4d793ba2f80 (Metraipolis archive dc_f4d793ba2f80, sha256 250cc8d225f47bfa72e8cb75a92be28573e9bb0875c4d8ce78eee8a8b756db50)
BibTeX
@misc{dcf4d793ba2f80,
title = {Labels Override Definitions in Jev-Style Typed Decision Models — arXiv:2610.02586},
howpublished = {arXiv},
url = {https://arxiv.org/abs/2610.02586},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f4d793ba2f80; sha256 250cc8d225f47bfa72e8cb75a92be28573e9bb0875c4d8ce78eee8a8b756db50},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.02586. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
A typed decision model answers a fixed question about an input by returning a probability for each of several caller-defined options. Each option carries a short label and a written definition, which is where a developer states the rule the model should apply. Jev introduced this interface for routing, moderation and triage, open implementations followed, and the same operation occurs whenever a language model is used as a classifier by scoring label strings. We study the open implementations, whose weights we can inspect and patch, and ask whether the probability follows the definitions or the labels. A preference for the label we call option-label bias. Across four open-weight typed decision models, three ways of reading an answer from a Qwen2.5 backbone, eleven classification tasks and PolicyBench, a synthetic routing suite we introduce in which the rule appears only in the definitions, the answer is mostly the labels. Deleting every definition leaves accuracy unchanged (laya-td: 0.8559 against 0.8487), although those definitions support 0.7971 on their own, and renaming the options to A and B raises accuracy by +0.1511 [+0.1377, +0.1646]. One system, von, is unaffected, and the two code bases differ in one expression: laya writes each option as "{label}: {definition}", while von writes only the definition. Changing that expression in both directions, with no weight changed, makes all three laya checkpoints exactly invariant (+0.0000 [+0.0000, +0.0000]) and creates the effect in von, whose accuracy falls from 0.8511 to 0.2281 when a label contradicts its definition. Earlier work attributed this failure to the constrained decision head these models use in place of a text decoder; our results locate it in the prompt rendering. We give a two-call test that tells a practitioner which case applies to their model, and measure what four mitigations are worth.
Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.02586 — 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 2026-10-05T22:45:55.603Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).