Instituto de Telecomunicações at IWSLT 2025: Aligning Small-Scale Speech and Language Models for Speech-to-Text Learning

Giuseppe Attanasio, Sonal Sannigrahi, Ben Peters, André F. T. Martins
6/20/2025
cs.CLcs.AI

No official code found. Best implementation: deep-spin/it-iwslt-2025 (2 stars, Apache-2.0, last commit 2025-09-30).

Abstract

This paper presents the IT-IST submission to the IWSLT 2025 Shared Task on Instruction Following Speech Processing. We submit results for the Short Track, i.e., speech recognition, translation, and spoken question answering. Our model is a unified speech-to-text model that integrates a pre-trained continuous speech encoder and text decoder through a first phase of modality alignment and a second phase of instruction fine-tuning. Crucially, we focus on using small-scale language model backbones (< 2B) and restrict to high-quality, CC-BY data along with synthetic data generation to supplement existing resources.

View on arXivView PDF

Code Implementations(1)

deep-spin/it-iwslt-2025Community62%
Apache-2.0

Code for the paper "Instituto de Telecomunicações at IWSLT 2025: Aligning Small-Scale Speech and Language Models for Speech-to-Text Learning"

20Shell, PythonMay 28, 20251 years agoApache-2.0

Similar Papers

KIT's Offline Speech Translation and Instruction Following Submission for IWSLT 2025

83% match2025

DeSTA2: Developing Instruction-Following Speech Language Model Without Speech Instruction-Tuning Data

81% match1 repo1272024

VoiceTextBlender: Augmenting Large Language Models with Speech Capabilities via Single-Stage Joint Speech-Text Supervised Fine-Tuning

81% match1 repo132024

Instruction Data Generation and Unsupervised Adaptation for Speech Language Models

81% match2024

Transferable speech-to-text large language model alignment module

80% match2024

Small-E: Small Language Model with Linear Attention for Efficient Speech Synthesis

80% match1 repo1772024

Cite this paper

@article{attanasio2025instituto,
  title  = {Instituto de Telecomunicações at IWSLT 2025: Aligning Small-Scale Speech and Language Models for Speech-to-Text Learning},
  author = {Giuseppe Attanasio and Sonal Sannigrahi and Ben Peters and André F. T. Martins},
  year   = {2025},
  eprint = {2506.17019},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2506.17019v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2506.17019.svg)](https://codeofpaper.com/paper/2506.17019)

Discussion