Isotropy Cliffs: The Geometric Signature of Decision-Making in Large Language Models

Okan S. Coskun; Florian Rottach; Carsten Eickhoff; William Rudman

Isotropy Cliffs: The Geometric Signature of Decision-Making in Large Language Models

Artificial Intelligence 2026-08-01 v1

Authors: Okan S. Coskun , Florian Rottach , Carsten Eickhoff , William Rudman

Abstract

We investigate the geometry of decision-making in Multiple Choice Question Answering (MCQA) through the lens of isotropy. Analyzing five open-weight models across diverse datasets, we identify decision-critical transition layers characterized by a shift in isotropy, coinciding with a major representational change and the emergence of task-relevant clusters. We demonstrate that this synchronized geometric behavior is strongly correlated with downstream accuracy ( $r\approx0.84$ ), displaying its relevance for successful decision-making. Furthermore, we show that this transition is robust to prompt variations, suggesting that it reflects a general mechanism of model behavior.

Keywords

decision-making isotropy large language models multiple choice question answering geometric signature transition layers

Cite

@article{arxiv.2608.00828,
  title  = {Isotropy Cliffs: The Geometric Signature of Decision-Making in Large Language Models},
  author = {Okan S. Coskun and Florian Rottach and Carsten Eickhoff and William Rudman},
  journal= {arXiv preprint arXiv:2608.00828},
  year   = {2026}
}