SEE: Signal Embedding Energy for Quantifying Noise Interference in Large Audio Language Models

TL;DR

This paper introduces Signal Embedding Energy (SEE), a novel metric to quantify noise impact on Large Audio Language Models, revealing insights into robustness and guiding improved denoising strategies for real-world applications.

Contribution

The paper proposes SEE, a new quantitative measure based on internal model representations, to assess and improve noise robustness in Large Audio Language Models.

Findings

SEE correlates strongly (0.98) with LALM performance.

Traditional denoising methods are often ineffective or harmful for LALMs.

SEE-based denoising strategies outperform existing methods.

Abstract

Large Audio Language Models (LALMs) have been widely applied in real-time scenarios, such as in-car assistants and online meeting comprehension. In practice, audio inputs are often corrupted by device and environmental noise, leading to performance degradation. However, existing LALM studies on noise lack quantitative analysis and rely mainly on intuition and empirical observation, thus failing to understand practical robustness. To address this issue, we introduce Signal Embedding Energy (SEE), a method for quantifying the impact of noise intensity on LALM inputs, enabling the differentiation of LALM robustness in real-world deployments. SEE introduces a perspective based on structured activation subspaces derived from the model's internal representations, which more accurately captures its perception of noise than raw audio features. Across experiments, SEE exhibits a strong correlation with LALM performance, achieving a correlation of 0.98. Surprisingly, traditional audio denoising methods are only marginally effective for LALMs, and, in some cases, even increase SEE and impair performance. This suggests a mismatch between speech-centric denoising objectives and the noise sensitivity of modern LALMs. Therefore, we propose a mitigation strategy derived from SEE to denoise LALM inputs, outperforming existing denoising methods. This paper introduces a novel metric for noise quantification in LALMs, providing guidance for robustness improvements in real-world deployments.