Do Bias Benchmarks Generalise? Evidence from Voice-based Evaluation of Gender Bias in SpeechLLMs

• URL: http://arxiv.org/abs/2510.01254v1
• Date: Wed, 24 Sep 2025 16:16:59 GMT
• Title: Do Bias Benchmarks Generalise? Evidence from Voice-based Evaluation of Gender Bias in SpeechLLMs
• Abstract summary: We fine-tune three SpeechLLMs using LoRA adapters to induce specific MCQA behaviours.<n>We evaluate whether these behaviours generalise to another, distinct MCQA benchmark, and more critically to long-form, creative generation tasks.
• Score: 22.24010339935349
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Recent work in benchmarking bias and fairness in speech large language models (SpeechLLMs) has relied heavily on multiple-choice question answering (MCQA) formats. The model is tasked to choose between stereotypical, anti-stereotypical, or neutral/irrelevant answers given an input speech prompt and an optional text prompt. Such MCQA benchmarks implicitly assume that model performance is consistent across other MCQA tasks, voices, and other task formats such as more realistic, long-form evaluations. In this paper, we probe that assumption. We fine-tune three SpeechLLMs using LoRA adapters to induce specific MCQA behaviours: preference for stereotypical, anti-stereotypical, or neutral/uncertain answers. We then evaluate whether these behaviours generalise to another, distinct MCQA benchmark, and more critically to long-form, creative generation tasks. Our results show that performance on MCQA bias benchmarks fails to reliably predict performances across other MCQA benchmarks, and more importantly across long-form tasks. We conclude that current MCQA bias benchmarks show limited evidence of cross-task generalisation in the speech domain, and also propose an evaluation suite for measuring behaviour transferability in future models and benchmarks.

Related papers

• Auditing MCQA Benchmarks through Probability Landscapes [15.39409235702177]
  We propose a two-component probabilistic framework for auditing multiple-choice question answering benchmarks.<n>For benchmark-level analysis, we characterize the probability landscape using the top prediction probability.<n>For item-level diagnostics, we introduce noise injection to reduce meaningful distractor competition.
  arXiv  Detail & Related papers  (2026-08-31T07:28:31Z)
• In-Situ Behavioral Evaluation for LLM Fairness, Not Standardized-Test Scores [20.623498601506974]
  We show that the standardized-test paradigm can be structurally unreliable.<n>We develop MAC-Fairness, a multi-agent conversational framework that embeds controlled variation factors into multi-round dialogue.<n>In-situ behavioral evaluation reveals stable, model-specific behavioral signatures that could generalize across benchmarks differing in fairness targets and evaluation methodologies.
  arXiv  Detail & Related papers  (2026-04-21T18:38:50Z)
• AURA Score: A Metric For Holistic Audio Question Answering Evaluation [57.042210272137396]
  We introduce AQEval to enable systematic benchmarking of AQA metrics.<n>It is the first benchmark of its kind, consisting of 10k model responses annotated by multiple humans for their correctness and relevance.<n>Second, we conduct a comprehensive analysis of existing AQA metrics on AQEval, highlighting weak correlation with human judgment.<n>Third, we propose a new metric - AURA score, to better evaluate open-ended model responses.
  arXiv  Detail & Related papers  (2025-10-06T15:41:34Z)
• Benchmarking and Mitigating MCQA Selection Bias of Large Vision-Language Models [2.393011821499345]
  We investigate the presence and nature of selection bias in Large Vision-Language Models (LVLMs)<n>We propose an inference-time logit-level debiasing method that estimates an ensemble bias vector from general and contextual prompts.<n>Our method mitigates bias without retraining and is compatible with frozen LVLMs.
  arXiv  Detail & Related papers  (2025-09-20T20:45:47Z)
• AHELM: A Holistic Evaluation of Audio-Language Models [78.20477815156484]
  multimodal audio-language models (ALMs) take interleaved audio and text as input and output text.<n>AHELM is a benchmark that aggregates various datasets -- including 2 new synthetic audio-text datasets called PARADE and CoRe-Bench.<n>We also standardize the prompts, inference parameters, and evaluation metrics to ensure equitable comparisons across models.
  arXiv  Detail & Related papers  (2025-08-29T07:40:39Z)
• Reasoning Models are Test Exploiters: Rethinking Multiple-Choice [12.317748510370238]
  Large Language Models (LLMs) are asked to choose among a fixed set of choices.<n>Multiple-choice question-answering (McQCA) is a good proxy for the downstream performance of models.<n>This paper investigates the extent to which this trend continues to hold for state-of-the-art reasoning models.
  arXiv  Detail & Related papers  (2025-07-21T07:49:32Z)
• Right Answer, Wrong Score: Uncovering the Inconsistencies of LLM Evaluation in Multiple-Choice Question Answering [78.89231943329885]
  Multiple-Choice Question Answering (MCQA) is widely used to evaluate Large Language Models (LLMs)<n>We show that multiple factors can significantly impact the reported performance of LLMs.<n>We analyze whether existing answer extraction methods are aligned with human judgment.
  arXiv  Detail & Related papers  (2025-03-19T08:45:03Z)
• Addressing Blind Guessing: Calibration of Selection Bias in Multiple-Choice Question Answering by Video Language Models [16.252597615544317]
  Video Language Models (VLMs) are designed to answer complex video-focused questions.<n>Current benchmarks fail to capture the full reasoning capabilities of VLMs due to selection bias.<n>This study is the first focused investigation of selection bias in video-to-text LLM-powered models.
  arXiv  Detail & Related papers  (2024-10-18T07:52:22Z)
• Bias in Language Models: Beyond Trick Tests and Toward RUTEd Evaluation [49.3814117521631]
  Standard benchmarks of bias and fairness in large language models (LLMs) measure the association between the user attributes stated or implied by a prompt.<n>We develop analogous RUTEd evaluations from three contexts of real-world use: children's bedtime stories, user personas, and English language learning exercises.<n>We find that standard bias metrics have no significant correlation with the more realistic bias metrics.
  arXiv  Detail & Related papers  (2024-02-20T01:49:15Z)
• HANS, are you clever? Clever Hans Effect Analysis of Neural Systems [1.6267479602370545]
  Large Language Models (It-LLMs) have been exhibiting outstanding abilities to reason around cognitive states, intentions, and reactions of all people involved, letting humans guide and comprehend day-to-day social interactions effectively. Several multiple-choice questions (MCQ) benchmarks have been proposed to construct solid assessments of the models' abilities. However, earlier works are demonstrating the presence of inherent "order bias" in It-LLMs, posing challenges to the appropriate evaluation.
  arXiv  Detail & Related papers  (2023-09-21T20:52:18Z)
• RoMQA: A Benchmark for Robust, Multi-evidence, Multi-answer Question Answering [87.18962441714976]
  We introduce RoMQA, the first benchmark for robust, multi-evidence, multi-answer question answering (QA) We evaluate state-of-the-art large language models in zero-shot, few-shot, and fine-tuning settings, and find that RoMQA is challenging. Our results show that RoMQA is a challenging benchmark for large language models, and provides a quantifiable test to build more robust QA methods.
  arXiv  Detail & Related papers  (2022-10-25T21:39:36Z)
• A Theoretically Grounded Benchmark for Evaluating Machine Commonsense [6.725087407394836]
  Theoretically-answered Commonsense Reasoning (TG-CSR) is based on discriminative question answering, but with questions designed to evaluate diverse aspects of commonsense. TG-CSR is based on a subset of commonsense categories first proposed as a viable theory of commonsense by Gordon and Hobbs. Preliminary results suggest that the benchmark is challenging even for advanced language representation models designed for discriminative CSR question answering tasks.
  arXiv  Detail & Related papers  (2022-03-23T04:06:01Z)
• Utilizing Self-supervised Representations for MOS Prediction [51.09985767946843]
  Existing evaluations usually require clean references or parallel ground truth data. Subjective tests, on the other hand, do not need any additional clean or parallel data and correlates better to human perception. We develop an automatic evaluation approach that correlates well with human perception while not requiring ground truth data.
  arXiv  Detail & Related papers  (2021-04-07T09:44:36Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.