Metraipolis › archive › dc_91c1eca55b2a

Same Text, Different Prediction: Serving-Context Nondeterminism in Text Classifiers — arXiv:2610.09111

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.09111
License (as archived)
CC0-1.0
Archived
2026-10-08T11:26:01.323Z
SHA-256 of the archived text
ff76a0bfb6a7995a67ac38280ec65e0f44b812c168d33910335a6797e846b54f
Stable URL of this page
https://metraipolis.com/documento/dc_91c1eca55b2a

How to cite

Same Text, Different Prediction: Serving-Context Nondeterminism in Text Classifiers — arXiv:2610.09111 — arXiv (en), https://arxiv.org/abs/2610.09111 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_91c1eca55b2a (Metraipolis archive dc_91c1eca55b2a, sha256 ff76a0bfb6a7995a67ac38280ec65e0f44b812c168d33910335a6797e846b54f)
BibTeX
@misc{dc91c1eca55b2a,
  title = {Same Text, Different Prediction: Serving-Context Nondeterminism in Text Classifiers — arXiv:2610.09111},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.09111},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_91c1eca55b2a; sha256 ff76a0bfb6a7995a67ac38280ec65e0f44b812c168d33910335a6797e846b54f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.09111. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Deterministic inference is essential for reliable and trustworthy machine learning. Prior studies of text generation have shown that changing factors such as batch size, batch composition, hardware, or inference engine can alter the generated text, even when the prompt, model parameters, and sampling randomness are fixed. These differences have been attributed in part to floating-point non-associativity, shape-dependent kernel selection, and other implementation-level differences in numerical execution. However, it remains unclear whether, when, and to what extent the same factors affect text classification. We present a systematic study of serving-context non-invariance in text classifiers, which prior work has measured only through generated text. We train 180 models spanning discriminative, pseudo-generative, and fully generative classifier formulations and evaluate each across four categories of serving contexts, holding the checkpoint and the text fixed. Label stability does not imply score stability. Changing only the batch shape changes no labels across fp32 comparisons, yet under bf16 it moves up to 56.7 percentage points of predicted probability mass, with label changes concentrated at small margins. Fully generative classifiers change more labels than their discriminative counterparts under the same serving changes. We derive sufficient conditions for label stability under each serving change and give a separate mitigation for each mechanism. Our results identify and quantify the serving conditions that must be fixed for reproducible text classification.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.09111 — 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 2026-10-08T10:24:06.310Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).