Benchmarking Time-localized Explanations for Audio Classification Models

2025-06-04

AI Summary

Audio models suffer from poor decision interpretability, and a lack of temporally grounded ground-truth annotations hinders objective evaluation of explanation quality. To address this, we introduce the first time-localized explanation benchmark for audio classification, proposing—novelly—a framework that leverages event-level temporal annotations as proxy ground truth. This enables reproducible, model-agnostic, quantitative evaluation of post-hoc temporal explanation methods (e.g., time-adapted Grad-CAM, SHAP, LIME). We design time-aligned evaluation metrics and a standardized dataset curation protocol to expose spurious temporal correlations. Our benchmark achieves an average IoU of >0.92 across multiple tasks—substantially outperforming baselines—and successfully identifies temporal pseudo-correlations embedded in training data.

Technology Category

• Machine Learning: Transparent, Interpretable, Explainable ML
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Computer Vision: Interpretability, Explainability, and Transparency

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Most modern approaches for audio processing are opaque, in the sense that they do not provide an explanation for their decisions. For this reason, various methods have been proposed to explain the outputs generated by these models. Good explanations can result in interesting insights about the data or the model, as well as increase trust in the system. Unfortunately, evaluating the quality of explanations is far from trivial since, for most tasks, there is no clear ground truth explanation to use as reference. In this work, we propose a benchmark for time-localized explanations for audio classification models that uses time annotations of target events as a proxy for ground truth explanations. We use this benchmark to systematically optimize and compare various approaches for model-agnostic post-hoc explanation, obtaining, in some cases, close to perfect explanations. Finally, we illustrate the utility of the explanations for uncovering spurious correlations.

Problem

• Evaluating quality of explanations for audio classification models
• Proposing benchmark using time annotations as ground truth
• Uncovering spurious correlations via optimized explanation methods

Innovation

• Benchmarking time-localized audio model explanations
• Using time annotations as ground truth proxy
• Systematically optimizing model-agnostic post-hoc explanations