Rare Word Recognition and Translation Without Fine-Tuning via Task Vector in Speech Models

TL;DR

This paper introduces a training-free method using task vectors for recognizing and translating rare words in speech models, achieving comparable or better results than fine-tuning while reducing costs and avoiding catastrophic forgetting.

Contribution

The authors propose a novel task vector approach that enables scalable, flexible, and training-free rare word recognition and translation in speech models.

Findings

Matches or surpasses fine-tuned models on target words

Improves general performance by about 5 BLEU

Mitigates catastrophic forgetting

Abstract

Rare words remain a critical bottleneck for speech-to-text systems. While direct fine-tuning improves recognition of target words, it often incurs high cost, catastrophic forgetting, and limited scalability. To address these challenges, we propose a training-free paradigm based on task vectors for rare word recognition and translation. By defining task vectors as parameter differences and introducing word-level task vector arithmetic, our approach enables flexible composition of rare-word capabilities, greatly enhancing scalability and reusability. Extensive experiments across multiple domains show that the proposed method matches or surpasses fine-tuned models on target words, improves general performance by about 5 BLEU, and mitigates catastrophic forgetting.