Robust ASR under domain shift is crucial because real-world systems encounter unseen accents and domains with limited labeled data. Although pseudo-labeling offers a practical workaround, it often introduces systematic, accent-specific errors that filtering fails to fix. We ask: How can we correct these recurring biases without target ground truth? We propose a simple parameter-space correction: in a source domain containing both real and pseudo-labeled data, two ASR models are fine-tuned from the same initialization, one on ground-truth labels and the other on pseudo-labels, and their weight difference forms a correction vector that captures pseudo-label biases. When applied to a pseudo-labeled target model, this vector enhances recognition, achieving up to a 35% relative Word Error Rate (WER) reduction on AfriSpeech-200 across ten African accents with the Whisper tiny model.
@article{arxiv.2510.08047,
title = {Pseudo2Real: Task Arithmetic for Pseudo-Label Correction in Automatic Speech Recognition},
author = {Yi-Cheng Lin and Yu-Hsuan Li Liang and Hsuan Su and Tzu-Quan Lin and Shang-Tse Chen and Yun-Nung Chen and Hung-yi Lee},
journal= {arXiv preprint arXiv:2510.08047},
year = {2026}
}