Training-Free Spectral Fingerprints of Voice Processing in Transformers

TL;DR

This paper introduces a training-free spectral analysis method to detect architectural signatures and biases in transformer models' voice processing, revealing how training emphasis influences connectivity patterns during linguistic tasks.

Contribution

It presents a novel spectral fingerprinting framework using graph signal processing to analyze transformer attention structures without training, uncovering model-specific and language-specific signatures.

Findings

Phi-3-Mini shows English-specific early layer disruption.

Spectral signatures correlate strongly with behavioral differences.

Attention head ablations confirm the functional relevance of spectral effects.

Abstract

Different transformer architectures implement identical linguistic computations via distinct connectivity patterns, yielding model imprinted ``computational fingerprints'' detectable through spectral analysis. Using graph signal processing on attention induced token graphs, we track changes in algebraic connectivity (Fiedler value, $\Delta\lambda_2$) under voice alternation across 20 languages and three model families, with a prespecified early window (layers 2--5). Our analysis uncovers clear architectural signatures: Phi-3-Mini shows a dramatic English specific early layer disruption ($\overline{\Delta\lambda_2}_{[2,5]}\!\approx\!-0.446$) while effects in 19 other languages are minimal, consistent with public documentation that positions the model primarily for English use. Qwen2.5-7B displays small, distributed shifts that are largest for morphologically rich languages, and LLaMA-3.2-1B exhibits systematic but muted responses. These spectral signatures correlate strongly with behavioral differences (Phi-3: $r=-0.976$) and are modulated by targeted attention head ablations, linking the effect to early attention structure and confirming functional relevance. Taken together, the findings are consistent with the view that training emphasis can leave detectable computational imprints: specialized processing strategies that manifest as measurable connectivity patterns during syntactic transformations. Beyond voice alternation, the framework differentiates reasoning modes, indicating utility as a simple, training free diagnostic for revealing architectural biases and supporting model reliability analysis.