An Analysis of Residual-Stream Geometry Across Transformer Depth
Not on the map. Why?
Abstract (arXiv)
We propose a transition-centred geometric analysis of transformer residual streams. Relative displacement measures how far representations move between consecutive layers, and orthogonal Procrustes analysis separates each transition into a rigid rotation and a non-rigid residual. Across six instruction-tuned models, on code generation and cross-lingual translation, these measurements reveal reproducible depth regularities. Relative displacement is strongly layer-dependent; typically larger early and late, with a quieter middle third; and nearly invariant across conditions within each model. Rotation magnitude is nearly constant across depth, while Procrustes residual and angle concentration remain depth-modulated, with residual peaking at the final transition. During generation, non-English targets show larger final-layer displacement and residual than English targets. We present these as descriptive geometric regularities, not as measures of computational effort or causal explanations. The contribution is a measurement framework for residual-stream transitions and evidence that, in the settings studied here, depth curves are model-dependent and largely condition-stable.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 19 of its 29 references that are in the map. First posted on arXiv 20 Jul 2026.
Builds on
19 of 29 references.
References · 19
- 2017Attention Is All You Needcited by 59,932
- 2024LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Codecited by 1,906
- 2019Similarity of Neural Network Representations Revisitedcited by 1,695
- 2023Progress measures for grokking via mechanistic interpretabilitycited by 822
- 2023Eliciting Latent Predictions from Transformers with the Tuned Lenscited by 579
- 2024Do Llamas Work in English? On the Latent Language of Multilingual Transformerscited by 275
- 2025Reasoning with Latent Thoughts: On the Power of Looped Transformerscited by 223
- 2019Emerging Cross-lingual Structure in Pretrained Language Modelscited by 180
- 2022The Geometry of Multilingual Language Model Representationscited by 94
- 2025Do Multilingual LLMs Think In English?cited by 75
- 2025Do Language Models Use Their Depth Efficiently?cited by 48
- 2025The Geometry of Reasoning: Flowing Logics in Representation Spacecited by 21
7 more
- 2025The Intrinsic Dimension of Prompts in Internal Representations of Large Language Modelscited by 17
- 2023Unveiling Multilinguality in Transformer Models: Exploring Language Specificity in Feed-Forward Networkscited by 12
- 2026Truth as a Trajectory: What Internal Representations Reveal About Large Language Model Reasoningcited by 9
- 2024The Local Interaction Basis: Identifying Computationally-Relevant and Sparsely Interacting Features in Neural Networkscited by 8
- 2026Context Structure Reshapes the Representational Geometry of Language Modelscited by 6
- 2025Unraveling the geometry of visual relational reasoningcited by 3
- 2026Shared Doubt: Zero-Shot Cross-Lingual Confidence Estimation for Language Modelscited by 1
Built on by
Citing papers per year: 2026: 1.
1 paper.