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

• URL: http://arxiv.org/abs/2608.00828v1
• Date: Sat, 01 Aug 2026 19:11:10 GMT
• Title: Isotropy Cliffs: The Geometric Signature of Decision-Making in Large Language Models
• Abstract summary: We investigate the geometry of decision-making in Multiple Choice Question Answering (MCQA) through the lens of isotropy.<n>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.
• Score: 22.87328932419285
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• 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.

Related papers

• Nonperturbative features in the Lie-algebraic Kähler sigma model with fermions [0.0]
  We investigate a quantum mechanical system originating from a Lie-algebraic K"ahler sigma model with multiple right-handed chiral fermions.<n>We identify and analyze saddle point solutions and examine their contributions within the perturbative expansions of the ground state energy.<n>We propose that the elongation parameter becomes relevant in shaping the system's quantum behavior from the three-loop level.
  arXiv  Detail & Related papers  (2024-12-16T04:55:14Z)
• Relative Representations: Topological and Geometric Perspectives [50.85040046976025]
  Relative representations are an established approach to zero-shot model stitching.<n>We introduce a normalization procedure in the relative transformation, resulting in invariance to non-isotropic rescalings and permutations.<n>Second, we propose to deploy topological densification when fine-tuning relative representations, a topological regularization loss encouraging clustering within classes.
  arXiv  Detail & Related papers  (2024-09-17T08:09:22Z)
• Parameter identification in linear non-Gaussian causal models under general confounding [8.273471398838533]
  We study identification of the linear coefficients when such models contain latent variables. Our main result is a graphical criterion that is necessary and sufficient for deciding generic identifiability of direct causal effects. We report on estimations based on the identification result, explore a generalization to models with feedback loops, and provide new results on the identifiability of the causal graph.
  arXiv  Detail & Related papers  (2024-05-31T14:39:14Z)
• Machine Learning for the identification of phase-transitions in interacting agent-based systems: a Desai-Zwanzig example [0.0]
  We propose a data-driven framework that pinpoints phase transitions for an agent-based model in its mean-field limit. To this end, we use the manifold learning algorithm Maps to identify a parsimonious set of data-driven latent variables. We then utilize a deep learning framework to obtain a conformal reparametrization of the data-driven coordinates.
  arXiv  Detail & Related papers  (2023-10-29T15:07:08Z)
• Exact solution for quantum strong long-range models via a generalized Hubbard-Stratonovich transformation [0.0]
  We present an exact analytical solution for quantum strong long-range models in the canonical ensemble. We utilize the equivalence between generalized Dicke models and interacting quantum models as a generalization of the Hubbard-Stratonovich transformation.
  arXiv  Detail & Related papers  (2023-05-17T18:00:02Z)
• EqMotion: Equivariant Multi-agent Motion Prediction with Invariant Interaction Reasoning [83.11657818251447]
  We propose EqMotion, an efficient equivariant motion prediction model with invariant interaction reasoning. We conduct experiments for the proposed model on four distinct scenarios: particle dynamics, molecule dynamics, human skeleton motion prediction and pedestrian trajectory prediction. Our method achieves state-of-the-art prediction performances on all the four tasks, improving by 24.0/30.1/8.6/9.2%.
  arXiv  Detail & Related papers  (2023-03-20T05:23:46Z)
• Towards Robust and Adaptive Motion Forecasting: A Causal Representation Perspective [72.55093886515824]
  We introduce a causal formalism of motion forecasting, which casts the problem as a dynamic process with three groups of latent variables. We devise a modular architecture that factorizes the representations of invariant mechanisms and style confounders to approximate a causal graph. Experiment results on synthetic and real datasets show that our three proposed components significantly improve the robustness and reusability of the learned motion representations.
  arXiv  Detail & Related papers  (2021-11-29T18:59:09Z)
• A Variational Inference Approach to Inverse Problems with Gamma Hyperpriors [60.489902135153415]
  This paper introduces a variational iterative alternating scheme for hierarchical inverse problems with gamma hyperpriors. The proposed variational inference approach yields accurate reconstruction, provides meaningful uncertainty quantification, and is easy to implement.
  arXiv  Detail & Related papers  (2021-11-26T06:33:29Z)
• Quantitative analysis of phase transitions in two-dimensional XY models using persistent homology [0.0]
  We use persistent homology and persistence images as an observable of three different variants of the two-dimensional XY model. For each model we successfully identify its phase transition(s) and are able to get an accurate determination of the critical temperatures and critical exponents of the correlation length.
  arXiv  Detail & Related papers  (2021-09-22T18:24:54Z)
• Semantic Change Detection with Asymmetric Siamese Networks [71.28665116793138]
  Given two aerial images, semantic change detection aims to locate the land-cover variations and identify their change types with pixel-wise boundaries. This problem is vital in many earth vision related tasks, such as precise urban planning and natural resource management. We present an asymmetric siamese network (ASN) to locate and identify semantic changes through feature pairs obtained from modules of widely different structures.
  arXiv  Detail & Related papers  (2020-10-12T13:26:30Z)
• Estimation of Switched Markov Polynomial NARX models [75.91002178647165]
  We identify a class of models for hybrid dynamical systems characterized by nonlinear autoregressive (NARX) components. The proposed approach is demonstrated on a SMNARX problem composed by three nonlinear sub-models with specific regressors.
  arXiv  Detail & Related papers  (2020-09-29T15:00:47Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.