# Automated Gradient-Driven Parameter Sharing for Low-Resource Multilingual Speech-to-Text Translation

arXiv: https://arxiv.org/abs/2603.25836 · HTML: https://mlsanity.com/paper/2603.25836/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 10 of its 15 references that are in the map.
- First posted on arXiv 26 Mar 2026.

## Builds on (most cited first, 10 of 10)

- [Gradient Surgery for Multi-Task Learning](https://mlsanity.com/paper/2001.06782/) (2020)
- [Domain Separation Networks](https://mlsanity.com/paper/1608.06019/) (2016)
- [BIG-C: a Multimodal Multi-Purpose Dataset for Bemba](https://mlsanity.com/paper/2305.17202/) (2023)
- [Disentangling Task Conflicts in Multi-Task LoRA via Orthogonal Gradient Projection](https://mlsanity.com/paper/2601.09684/) (2026)
- [GDOD: Effective Gradient Descent using Orthogonal Decomposition for Multi-Task Learning](https://mlsanity.com/paper/2301.13465/) (2023)
- [Canonical Correlation Analysis in high dimensions with structured regularization](https://mlsanity.com/paper/2011.01650/) (2020)
- [Attention Is Not All You Need: The Importance of Feedforward Networks in Transformer Models](https://mlsanity.com/paper/2505.06633/) (2025)
- [Finetuning End-to-End Models for Estonian Conversational Spoken Language Translation](https://mlsanity.com/paper/2407.03809/) (2024)
- [Gradient Deconfliction via Orthogonal Projections onto Subspaces For Multi-task Learning](https://mlsanity.com/paper/2503.03438/) (2025)
- [GMU Systems for the IWSLT 2025 Low-Resource Speech Translation Shared Task](https://mlsanity.com/paper/2505.21781/) (2025)

## Abstract (arXiv)

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.