Pruning for Efficiency, Paying in Fairness: Demographic Disparities in Pruned Speech-LLMs
cs.SD, cs.CL
Submitted: 2026-09-29
Updated: 2026-09-29
Code: https://github.com/KarthikKolluriKB/SpeechLLM-Pruning-Fairness
Terminology
Sources
- Common Voice: A Massively-Multilingual Speech Corpus
- Kid-Whisper: Towards Bridging the Performance Gap in Automatic Speech Recognition for Children VS. Adults
- On Fairness of Low-Rank Adaptation of Large Models
- Quantifying Bias in Automatic Speech Recognition
- Do LLM Decoders Listen Fairly? Benchmarking How Language Model Priors Shape Bias in Speech Recognition
- Understanding the Effect of Model Compression on Social Bias in Large Language Models
- Characterising Bias in Compressed Models
- LoRA: Low-Rank Adaptation of Large Language Models
- On the social bias of speech self-supervised models
- An Embarrassingly Simple Approach for LLM with Strong ASR Capacity
- Qwen2.5 Technical Report
- Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization
- Improving the Inclusivity of Dutch Speech Recognition by Fine-tuning Whisper on the JASMIN-CGN Corpus
- Pruning has a disparate impact on model accuracy
- Towards measuring fairness in speech recognition: Fair-Speech dataset
Related papers
- Few-Shot Open-Set Audio Classification via Transductive Prototype Refinement and Class Logit Enhancement
- Spectral Masking and Interpolation Attack (SMIA): A Black-box Adversarial Attack against Voice Authentication and Anti-Spoofing Systems
- AVMeme Exam: A Multimodal Multilingual Multicultural Benchmark for LLMs' Contextual and Cultural Knowledge and Thinking
- SoundWeaver: Compositional Warm-Starting for Text-to-Audio Diffusion Serving
- WASIL: In-the-Wild Arabic Spoken Interactions with LLMs
- Efficient Audiovisual Speech Processing via MUTUD: Multimodal Training and Unimodal Deployment