Depth-Wise Emergence of Prediction-Centric Geometry in Large Language Models

📅 2026-02-04
📈 Citations: 0
Influential: 0
📄 PDF
🤖 AI Summary
This work investigates how decoder-only large language models transform contextual information into predictive outputs along the depth dimension. By integrating geometric representation analysis with mechanistic interventions, the study reveals—for the first time—that angular components of deep-layer representations encode similarity structures aligned with predictive distributions, while norm components carry non-predictive contextual information. This finding establishes a mechanism-geometric account of the context-to-prediction transformation process. Building upon disentangled representations and an intervenable modeling framework, the research identifies structured geometric properties underlying prediction formation and demonstrates causal, selective control over token-level predictions.

Technology Category

Application Category

📝 Abstract
We show that decoder-only large language models exhibit a depth-wise transition from context-processing to prediction-forming phases of computation accompanied by a reorganization of representational geometry. Using a unified framework combining geometric analysis with mechanistic intervention, we demonstrate that late-layer representations implement a structured geometric code that enables selective causal control over token prediction. Specifically, angular organization of the representation geometry parametrizes prediction distributional similarity, while representation norms encode context-specific information that does not determine prediction. Together, these results provide a mechanistic-geometric account of the dynamics of transforming context into predictions in LLMs.
Problem

Research questions and friction points this paper is trying to address.

large language models
representational geometry
prediction
depth-wise dynamics
context-to-prediction transformation
Innovation

Methods, ideas, or system contributions that make the work stand out.

representational geometry
depth-wise dynamics
prediction-centric coding
mechanistic interpretability
decoder-only LLMs
S
Shahar Haim
Champalimaud Centre for the Unknown, Lisbon, Portugal
D
Daniel C. McNamee
Champalimaud Centre for the Unknown, Lisbon, Portugal