Saved in:
Bibliographic Details
Main Authors: Sun, Ruiyan, Nakamura, Satoshi
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.25836
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914425840074752
author Sun, Ruiyan
Nakamura, Satoshi
author_facet Sun, Ruiyan
Nakamura, Satoshi
contents In low-resource multilingual speech-to-text translation, uniform architectural sharing across languages frequently introduces representation conflicts that impede convergence. This work proposes a principled methodology to automatically determine layer-specific sharing patterns by mining training gradient information. Our approach employs three distinct analysis strategies: distance-based language clustering, self/cross-task divergence metrics for capacity allocation, and joint factorization coupled with canonical correlation analysis for subspace alignment. Extensive evaluation across four language pairs (using the SeamlessM4T-Medium architecture) demonstrates persistent improvements in translation quality metrics.
format Preprint
id arxiv_https___arxiv_org_abs_2603_25836
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Gradient-Informed Training for Low-Resource Multilingual Speech Translation
Sun, Ruiyan
Nakamura, Satoshi
Computation and Language
In low-resource multilingual speech-to-text translation, uniform architectural sharing across languages frequently introduces representation conflicts that impede convergence. This work proposes a principled methodology to automatically determine layer-specific sharing patterns by mining training gradient information. Our approach employs three distinct analysis strategies: distance-based language clustering, self/cross-task divergence metrics for capacity allocation, and joint factorization coupled with canonical correlation analysis for subspace alignment. Extensive evaluation across four language pairs (using the SeamlessM4T-Medium architecture) demonstrates persistent improvements in translation quality metrics.
title Gradient-Informed Training for Low-Resource Multilingual Speech Translation
topic Computation and Language
url https://arxiv.org/abs/2603.25836