Saved in:
Bibliographic Details
Main Authors: Pang, Zi Haur, Gao, Xiaoxue, Kawahara, Tatsuya, Chen, Nancy F.
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.21050
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914413500432384
author Pang, Zi Haur
Gao, Xiaoxue
Kawahara, Tatsuya
Chen, Nancy F.
author_facet Pang, Zi Haur
Gao, Xiaoxue
Kawahara, Tatsuya
Chen, Nancy F.
contents Speech emotion recognition (SER) systems can exhibit gender-related performance disparities, but how such bias manifests in multilingual speech LLMs across languages and modalities is unclear. We introduce a novel multilingual, multimodal benchmark built on MELD-ST, spanning English, Japanese, and German, to quantify language-specific SER performance and gender gaps. We find bias is strongly language-dependent, and multimodal fusion does not reliably improve fairness. To address these, we propose ERM-MinMaxGAP, a fairness-informed training objective, which augments empirical risk minimization (ERM) with a proposed adaptive fairness weight mechanism and a novel MinMaxGAP regularizer on the maximum male-female loss gap within each language and modality. Building upon the Qwen2-Audio backbone, our ERM-MinMaxGAP approach improves multilingual SER performance by 5.5% and 5.0% while reducing the overall gender bias gap by 0.1% and 1.4% in the unimodal and multimodal settings, respectively.
format Preprint
id arxiv_https___arxiv_org_abs_2603_21050
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle ERM-MinMaxGAP: Benchmarking and Mitigating Gender Bias in Multilingual Multimodal Speech-LLM Emotion Recognition
Pang, Zi Haur
Gao, Xiaoxue
Kawahara, Tatsuya
Chen, Nancy F.
Sound
Speech emotion recognition (SER) systems can exhibit gender-related performance disparities, but how such bias manifests in multilingual speech LLMs across languages and modalities is unclear. We introduce a novel multilingual, multimodal benchmark built on MELD-ST, spanning English, Japanese, and German, to quantify language-specific SER performance and gender gaps. We find bias is strongly language-dependent, and multimodal fusion does not reliably improve fairness. To address these, we propose ERM-MinMaxGAP, a fairness-informed training objective, which augments empirical risk minimization (ERM) with a proposed adaptive fairness weight mechanism and a novel MinMaxGAP regularizer on the maximum male-female loss gap within each language and modality. Building upon the Qwen2-Audio backbone, our ERM-MinMaxGAP approach improves multilingual SER performance by 5.5% and 5.0% while reducing the overall gender bias gap by 0.1% and 1.4% in the unimodal and multimodal settings, respectively.
title ERM-MinMaxGAP: Benchmarking and Mitigating Gender Bias in Multilingual Multimodal Speech-LLM Emotion Recognition
topic Sound
url https://arxiv.org/abs/2603.21050