MLLP-VRAIN UPV system for the IWSLT 2025 Simultaneous Speech Translation Translation task

TL;DR

This paper presents a modular, real-time speech translation system for long-form content that combines pre-trained models with adaptive strategies to balance translation quality and latency in the IWSLT 2025 challenge.

Contribution

It introduces a novel cascade system that adapts strong pre-trained models for streaming translation without extensive retraining, addressing long-form speech challenges.

Findings

Achieved BLEU score of 31.96 on ACL60/60 dataset.

Latency of 2.94 seconds with non-computational-aware StreamLAAL.

Preliminary test BLEU score of 29.8 on IWSLT25Instruct.

Abstract

This work describes the participation of the MLLP-VRAIN research group in the shared task of the IWSLT 2025 Simultaneous Speech Translation track. Our submission addresses the unique challenges of real-time translation of long-form speech by developing a modular cascade system that adapts strong pre-trained models to streaming scenarios. We combine Whisper Large-V3-Turbo for ASR with the multilingual NLLB-3.3B model for MT, implementing lightweight adaptation techniques rather than training new end-to-end models from scratch. Our approach employs document-level adaptation with prefix training to enhance the MT model's ability to handle incomplete inputs, while incorporating adaptive emission policies including a wait-$k$ strategy and RALCP for managing the translation stream. Specialized buffer management techniques and segmentation strategies ensure coherent translations across long audio sequences. Experimental results on the ACL60/60 dataset demonstrate that our system achieves a favorable balance between translation quality and latency, with a BLEU score of 31.96 and non-computational-aware StreamLAAL latency of 2.94 seconds. Our final model achieves a preliminary score on the official test set (IWSLT25Instruct) of 29.8 BLEU. Our work demonstrates that carefully adapted pre-trained components can create effective simultaneous translation systems for long-form content without requiring extensive in-domain parallel data or specialized end-to-end training.