# Temporal Taxation Compounds Under Post-Training Compression of Whisper Models

arXiv: https://arxiv.org/abs/2609.28739 · HTML: https://mlsanity.com/paper/2609.28739/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 18 of its 31 references that are in the map.
- First posted on arXiv 23 Sep 2026.

## Builds on (most cited first, 18 of 18)

- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers](https://mlsanity.com/paper/2210.17323/) (2022)
- [SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot](https://mlsanity.com/paper/2301.00774/) (2023)
- [LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale](https://mlsanity.com/paper/2208.07339/) (2022)
- [What Do Compressed Deep Neural Networks Forget?](https://mlsanity.com/paper/1911.05248/) (2019)
- [Characterising Bias in Compressed Models](https://mlsanity.com/paper/2010.03058/) (2020)
- [Distil-Whisper: Robust Knowledge Distillation via Large-Scale Pseudo Labelling](https://mlsanity.com/paper/2311.00430/) (2023)
- [Careless Whisper: Speech-to-Text Hallucination Harms](https://mlsanity.com/paper/2402.08021/) (2024)
- [The Low-Resource Double Bind: An Empirical Study of Pruning for Low-Resource Machine Translation](https://mlsanity.com/paper/2110.03036/) (2021)
- [On The Impact of Machine Learning Randomness on Group Fairness](https://mlsanity.com/paper/2307.04138/) (2023)
- [Intriguing Properties of Compression on Multilingual Models](https://mlsanity.com/paper/2211.02738/) (2022)
- [The Impact of Inference Acceleration on Bias of LLMs](https://mlsanity.com/paper/2410.22118/) (2024)
- [Edge-ASR: Towards Low-Bit Quantization of Automatic Speech Recognition Models](https://mlsanity.com/paper/2507.07877/) (2025)
- [Fairness of Automatic Speech Recognition: Looking Through a Philosophical Lens](https://mlsanity.com/paper/2508.07143/) (2025)
- [Do LLM Decoders Listen Fairly? Benchmarking How Language Model Priors Shape Bias in Speech Recognition](https://mlsanity.com/paper/2604.21276/) (2026)
- [Efficient Compression of Multitask Multilingual Speech Models](https://mlsanity.com/paper/2405.00966/) (2024)
- [Reexamining Racial Disparities in Automatic Speech Recognition Performance: The Role of Confounding by Provenance](https://mlsanity.com/paper/2407.13982/) (2024)
- [Beyond Word Error Rate: Auditing the Diversity Tax in Speech Recognition through Dataset Cartography](https://mlsanity.com/paper/2603.05267/) (2026)

## Abstract (arXiv)

Automatic speech recognition models are audited for demographic fairness at full precision, yet the models that ship to production have been quantized, pruned, and distilled. We ask whether post-training weight compression, which alters model weights rather than the audio signal or its feature representation, redistributes error burden across demographic groups. Across the Whisper family on Fair-Speech, Common Voice 25, and AfriSpeech-200, 50% Wanda pruning of Whisper-large-v3 sharply widens the Black/AA-vs-Asian temporal-taxation differential on Fair-Speech: the absolute word-error-rate gap between the worst- and best-served groups more than doubles; at an assumed cost of five seconds of correction effort per transcription error this is a rise from 30 to 64 seconds of correction time per minute of speech. This +111% relative increase is invariant to the assumed per-error cost, survives an audio-quality control, and is only partly mitigated by beam-search decoding, which still leaves an +86% increase. At edge model size, INT4 HQQ quantization compounds catastrophic transcript loops on West African accents by factors of five to seven. Distillation, by contrast, narrows demographic gaps in 21 of 27 evaluated settings (teacher-student pair, precision, and dataset), with the exceptions concentrated on a single model pair. We cast the temporal-taxation construct of Choi and Choi (2025) as a quantitative metric, and show that single-snapshot fairness audits on full-precision models do not capture the deployment-time burden that compression places on already-marginalized speakers.