paperSeptember 2026Unreviewed

Consistency Without Alignment: Item-Sensitive Language Models Indistinguishable From Random

Cris Huynh

Abstract

Item-sensitivity, defined as whether a model's choice depends on the specific input rather than on its own output prior, is widely reported as evidence of task competence. We show this evidence is necessary but not sufficient using a forced-choice signalling task abstracted from the board game Deception: Murder in Hong Kong. In this environment, the reference points against which a coordinate should be judged (a fit-maximising strategy, a posterior-maximising strategy, and uniform random selecti

Cite

@misc{huynh2026consistency,
  title = {{Consistency Without Alignment: Item-Sensitive Language Models Indistinguishable From Random}},
  author = {Cris Huynh},
  year = {2026},
  month = sep,
  eprint = {2609.00576},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.00576}
}