Representational Similarity Analysis (RSA)

methods / Representation Alignment / Representational Similarity Analysis (RSA)

Techniqueintermediate

Compares two representations of the same items by building one pairwise-distance matrix per representation and correlating the two matrices (e.g. via Kendall's tau), rather than comparing raw coordinates directly — the classical, originally-neuroscientific alignment technique this map's other representation-alignment methods refine.

Used in (14 observations)

Sinusoidal number geometry recurs across eight LLMs and is causalmeasured

structure: Lissajous Curves · models: OLMo 2 1B, OLMo 2 7B, OLMo 2 13B, Llama-3.2-1B, Llama-3.2-3B, Llama-3-8B, Llama-3.1-8B, Phi-4 (15B) · paper: Unravelling the Mechanisms of Manipulating Numbers in Language Models

Geographic representations are causally used for prediction, not just decodablemeasured

structure: Affine Subspace · models: DeBERTa-v2-xxlarge, GPT-Neo-1.3B · paper: More than Correlation: Do Large Language Models Learn Causal Representations of Space?

A GNN odor-map's geometry matches human perceptual odor similaritymeasured

structure: Platonic Representation Hypothesis, Linear Subspace · models: Principal Odor Map (Message Passing Neural Network) · paper: A Principal Odor Map Unifies Diverse Tasks in Human Olfactory Perception

LLM hidden states warp categorically at digit-count boundariesmeasured

structure: Decision boundary (as a codimension-1 hypersurface), Linear Direction · models: Llama-3-8B-Instruct, Mistral-7B-Instruct-v0.3, Gemma-2-9B-it, Qwen2.5-7B-Instruct, Phi-3.5-mini-instruct, Llama-3-8B · paper: Categorical Perception in Large Language Model Hidden States: Structural Warping at Digit-Count Boundaries

Human perceptual geometry emerges transiently in middle LLM layersmeasured

structure: Circle · models: Llama-3-8B, Llama-3.2-3B, Gemma-7B, Qwen3-4B · paper: Geometry of Human Perceptual Domains Emerges Transiently in LLM Representations

SAE feature spaces (not raw activations) trained independently on Pythia-70m/160m, Gemma-1/2, and Llama-3/3.1 show high SVCCA and RSA similarity at matched middle layers after activation-correlation feature pairing, far above random-pairing and randomized-weight-model baselines, and semantically matched concept subspaces (e.g. Emotions) are more similar across models than the feature space as a wholemeasured

structure: Platonic Representation Hypothesis · models: Pythia-70M, Pythia-160M, Gemma-2B, Gemma-2-2B, Gemma-2-9B, Llama-3-8B-Instruct, Llama-3.1-8B · paper: Quantifying Feature Space Universality Across Large Language Models via Sparse Autoencoders, Semantic Convergence: Investigating Shared Representations Across Scaled LLMs

Representational similarity analysis shows 15 language and vision-language models align more strongly with brain regions whose activity is consistent across sentence, word-cloud, and image presentations of the same concept than with less modality-consistent regionsmeasured

structure: Platonic Representation Hypothesis · models: BERT-base-uncased, GPT-2 Small, CLIP ViT-B/32, FLAVA · paper: Language models align with brain regions that represent concepts across modalities

In-context concept representations converge toward a human-aligned structuremeasured

structure: Platonic Representation Hypothesis, Linear Subspace · models: Llama-3-70B, Qwen2-0.5B · paper: Revealing Emergent Human-like Conceptual Representations from Language Prediction

Text-only LMs linearly recover CIELAB color-space structuremeasured

structure: Linear Subspace · models: BERT-large-uncased, RoBERTa-large, ELECTRA-large (discriminator), BERT-mini, BERT-small, BERT-medium, BERT-base-uncased · paper: Can Language Models Encode Perceptual Structure Without Grounding? A Case Study in Color

A pose-based sign model handshape geometry matches human perceptionmeasured

structure: Platonic Representation Hypothesis, Linear Subspace · models: I3D (fine-tuned on ASL Citizen, pixel-based), ST-GCN (fine-tuned on ASL Citizen, pose-based) · paper: Phonological Perception of Sign Language Models

RSA-localized concept vectors are surface-form invariant and causally guide behaviormeasured

structure: Linear Direction · models: Llama-3.1-8B, Llama 3.1 70B · paper: Analogical Reasoning Inside Large Language Models: Concept Vectors and the Limits of Abstraction

A Backpack model sense-vector topology partially transfers across languagesmeasured

structure: Linear Subspace · models: GPT-2-small, GPT-2-Medium, GPT-2-Large · paper: Multilinguality as Sense Adaptation

Music-task fine-tuning, not exposure, induces chroma equivalencemeasured

structure: Circle · models: Wav2Vec 2.0 Base (LibriSpeech-960h), Data2Vec Audio Base (LibriSpeech-960h) · paper: Musical Training, but not Mere Exposure to Music, Drives the Emergence of Chroma Equivalence in Artificial Neural Networks

Vision MoE-CNN experts partition into a stable animate/inanimate splitmeasured

structure: Concept Cluster Heterogeneity · models: Sparsely-gated contrastive MoE-CNN, E=8 experts (Tangtartharakul & Storrs 2026) · paper: Beyond Routing: Characterising Expert Tuning and Representation in Vision Mixture-of-Experts