Token Merging for Multilingual Speech Recognition: A Systematic Study Across Model Scale and Fine-Tuning

Author1, Author2, Author3, Author4, Author5

multilingualspeech recognitiontoken mergingfine-tuning

2609.13151

Builder Relevance

80%

2h ago

Abstract

Token merging improves computational efficiency in multilingual speech recognition with minimal impact on transcription accuracy.

Reality Card

Core Claim

Token merging significantly enhances computational efficiency in multilingual speech recognition models like Whisper without sacrificing transcription accuracy.

Method / Result

Token merging increases computational efficiency with almost no loss in transcription accuracy across most low-resource languages and model sizes.

Limitations

The study may not address the long-term effects of token merging on model performance in diverse real-world applications.

Paper to code

Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.

No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.

← Back to all papers