Training-Seed Variability in Speech LLM Adaptation
A study finds that training seed variability significantly impacts fairness metrics in speech LLMs, more so than audio compression factors.
01
Evidence
- AarXiv cs.CLInstitution2026-10-02 12:00
At 460 h of clean LibriSpeech, the seed moves fairness metrics more than compression does on most demographic axes.
- AarXiv cs.CLInstitution2026-10-02 12:00
We fine-tune the Q-former projector and LoRA adapters of a speech LLM at five audio compression factors and six random seeds
- AarXiv cs.CLInstitution2026-10-02 12:00
Abstract: Demographic fairness gaps in automatic speech recognition are almost always reported from a single training run. We fine-tune the Q-former projector…