# From Dispersion to Attraction: Spectral Dynamics of Hallucination Across Whisper Model Scales

arXiv: https://arxiv.org/abs/2604.08591 · HTML: https://mlsanity.com/paper/2604.08591/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 14 of its 14 references that are in the map.
- First posted on arXiv 31 Mar 2026.

## Builds on (most cited first, 14 of 14)

- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [Survey of Hallucination in Natural Language Generation](https://mlsanity.com/paper/2202.03629/) (2022)
- [A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions](https://mlsanity.com/paper/2311.05232/) (2023)
- [Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation](https://mlsanity.com/paper/2302.09664/) (2023)
- [Teaching Models to Express Their Uncertainty in Words](https://mlsanity.com/paper/2205.14334/) (2022)
- [Transformers need glasses! Information over-squashing in language tasks](https://mlsanity.com/paper/2406.04267/) (2024)
- [Careless Whisper: Speech-to-Text Hallucination Harms](https://mlsanity.com/paper/2402.08021/) (2024)
- [Tracing the Representation Geometry of Language Models from Pretraining to Post-training](https://mlsanity.com/paper/2509.23024/) (2025)
- [Investigation of Whisper ASR Hallucinations Induced by Non-Speech Audio](https://mlsanity.com/paper/2501.11378/) (2025)
- [Lost in Transcription, Found in Distribution Shift: Demystifying Hallucination in Speech Foundation Models](https://mlsanity.com/paper/2502.12414/) (2025)
- [Transformers Learn Low Sensitivity Functions: Investigations and Implications](https://mlsanity.com/paper/2403.06925/) (2024)
- [From Condensation to Rank Collapse: A Two-Stage Analysis of Transformer Training Dynamics](https://mlsanity.com/paper/2510.06954/) (2025)
- [Spectral Scaling Laws in Language Models: How Effectively Do Feed-Forward Networks Use Their Latent Space?](https://mlsanity.com/paper/2510.00537/) (2025)
- [Why Deep Jacobian Spectra Separate: Depth-Induced Scaling and Singular-Vector Alignment](https://mlsanity.com/paper/2602.12384/) (2026)

## Built on by (most cited first, 2 of 2)

- [The Null Token Knows: Reducing Message-Free Hallucination in ASR and NMT](https://mlsanity.com/paper/2608.15940/) (2026)
- [Repetition, Not Length: Isolating the Counting Failure in Neural Text-to-Speech](https://mlsanity.com/paper/2609.36974/) (2026)

## Abstract (arXiv)

Hallucinations in large ASR models present a critical safety risk. In this work, we propose the textitSpectral Sensitivity Theorem, which predicts a phase transition in deep networks from a dispersive regime (signal decay) to an attractor regime (rank-1 collapse) governed by layer-wise gain and alignment. We validate this theory by analyzing the eigenspectra of activation graphs in Whisper models (Tiny to Large-v3-Turbo) under adversarial stress. Our results confirm the theoretical prediction: intermediate models exhibit textitStructural Disintegration (Regime I), characterized by a 13.4% collapse in Cross-Attention rank. Conversely, large models enter a textitCompression-Seeking Attractor state (Regime II), where Self-Attention actively compresses rank (-2.34%) and hardens the spectral slope, decoupling the model from acoustic evidence.