Bias in the Ear of the Listener: Assessing Sensitivity in Audio Language Models Across Linguistic, Demographic, and Positional Variations

TL;DR

This paper introduces the BiasInEar dataset and evaluates multilingual speech language models, revealing their sensitivity to language and structural biases while being relatively robust to demographic factors.

Contribution

It presents the first systematic speech bias assessment framework and a new multilingual speech benchmark for evaluating fairness and robustness in MLLMs.

Findings

Models are sensitive to language and option order biases.

Models show robustness to demographic factors like gender.

Architectural design influences robustness across languages.

Abstract

This work presents the first systematic investigation of speech bias in multilingual MLLMs. We construct and release the BiasInEar dataset, a speech-augmented benchmark based on Global MMLU Lite, spanning English, Chinese, and Korean, balanced by gender and accent, and totaling 70.8 hours ($\approx$4,249 minutes) of speech with 11,200 questions. Using four complementary metrics (accuracy, entropy, APES, and Fleiss' $\kappa$), we evaluate nine representative models under linguistic (language and accent), demographic (gender), and structural (option order) perturbations. Our findings reveal that MLLMs are relatively robust to demographic factors but highly sensitive to language and option order, suggesting that speech can amplify existing structural biases. Moreover, architectural design and reasoning strategy substantially affect robustness across languages. Overall, this study establishes a unified framework for assessing fairness and robustness in speech-integrated LLMs, bridging the gap between text- and speech-based evaluation. The resources can be found at https://github.com/ntunlplab/BiasInEar.