bias mitigation 1checkpoint variability 1deep neural networks 1genre bias 1music evaluation 1self-supervised speech models 1shortcut learning 1speech classification 1SUPERB evaluation 1supervised fine-tuning 1
From the 2 of 3 linked papers with an AI index.
3 papers
cs.SD2026
Genre Bias or Aesthetic Perception? Identifying and Mitigating Shortcut Learning in Music Evaluation
Yizhou Zhang, Wangjin Zhou, Yi Zhao +3
The paper uncovers that music aesthetic scoring models often rely on genre cues as shortcuts, leading to biased evaluations, and introduces a training objective that reweights hard…
cs.SD2026
Rethinking Speech Foundation Model Fine-tuning: Better SFT or Better Match?
Wangjin Zhou, Yizhou Zhang, Yichi Wang +1
The paper investigates how supervised fine-tuning performance for speech foundation models varies across different pretrained checkpoints, showing that gains often depend on the sp…
cs.SD2026
SONAR: Self-Distilled Continual Pre-training for Domain Adaptive Audio Representation
Yizhou Zhang, Yuan Gao, Wangjin Zhou +3
Self-supervised learning (SSL) on large-scale datasets like AudioSet has become the dominant paradigm for audio representation learning. While the continuous influx of new, unlabel…