Human perceptual geometry emerges transiently in middle LLM layers

measured in 1 paper

Singh & Chopra extract per-layer last-token activations from four open LLMs (LLaMA-3-8B, LLaMA-3.2-3B, Gemma-7B, Qwen3-4B) for color, pitch, taste and emotion stimuli, with no probing or training [singh-chopra-2026-transient-perceptual-geometry] Each layer's 2D MDS qualia map is aligned to independent human ground truth via RSA (Spearman) and generalized Procrustes analysis, with bootstrapped 95% CIs over 1000 iterations [singh-chopra-2026-transient-perceptual-geometry] Color forms a circular manifold resembling the human color wheel, pitch a smooth arc, and emotion an affective manifold aligned with the valence-arousal structure [singh-chopra-2026-transient-perceptual-geometry] Alignment follows a rise-peak-fall trajectory across depth (weak early, strongest in intermediate layers, attenuated late), consistent across all four models and domains [singh-chopra-2026-transient-perceptual-geometry] Emotion is a partial exception, staying comparatively stable once peaked, while taste shows a noisier, faster-decaying trajectory [singh-chopra-2026-transient-perceptual-geometry] Exact per-layer RSA/GPA values live only in figures not extractable from the fetched source, so this entry uses the paper's quoted qualitative claims [singh-chopra-2026-transient-perceptual-geometry]

Structure

Circle

Context

qualia map (per-layer MDS projection), RSA and GPA alignment to human perceptual ground truth, transient (rise-peak-fall) layer-wise emergence, color wheel, pitch arc, valence-arousal manifold, taste geometry, fully intrinsic (no probing/training) methodology

Confirmed in models

Llama-3-8B · Llama, Llama-3.2-3B · Llama, Gemma-7B · Gemma, Qwen3-4B · Qwen

Method

Representational Similarity Analysis (RSA), Generalized Procrustes Analysis (GPA)

Papers

Geometry of Human Perceptual Domains Emerges Transiently in LLM Representations — Singh, Simardeep, Chopra, Paras2026 · arXiv:2605.27970