Cross-Attention is Half Explanation in Speech-to-Text Models

• URL: http://arxiv.org/abs/2509.18010v1
• Date: Mon, 22 Sep 2025 16:49:26 GMT
• Title: Cross-Attention is Half Explanation in Speech-to-Text Models
• Authors: Sara Papi, Dennis Fucci, Marco Gaido, Matteo Negri, Luisa Bentivogli,
• Abstract summary: Cross-attention is a core mechanism in encoder-decoder architectures, widespread in many fields, including speech-to-text (S2T) processing.<n>Our analysis spans monolingual and multilingual, single-task and multi-task models at multiple scales, and shows that attention scores moderately to strongly align with saliency-based explanations.<n>It also shows that cross-attention captures only about 50% of the input relevance and, in the best case, only partially reflects how the decoder attends to the encoder's representations.
• Score: 31.16674879591289
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Cross-attention is a core mechanism in encoder-decoder architectures, widespread in many fields, including speech-to-text (S2T) processing. Its scores have been repurposed for various downstream applications--such as timestamp estimation and audio-text alignment--under the assumption that they reflect the dependencies between input speech representation and the generated text. While the explanatory nature of attention mechanisms has been widely debated in the broader NLP literature, this assumption remains largely unexplored within the speech domain. To address this gap, we assess the explanatory power of cross-attention in S2T models by comparing its scores to input saliency maps derived from feature attribution. Our analysis spans monolingual and multilingual, single-task and multi-task models at multiple scales, and shows that attention scores moderately to strongly align with saliency-based explanations, particularly when aggregated across heads and layers. However, it also shows that cross-attention captures only about 50% of the input relevance and, in the best case, only partially reflects how the decoder attends to the encoder's representations--accounting for just 52-75% of the saliency. These findings uncover fundamental limitations in interpreting cross-attention as an explanatory proxy, suggesting that it offers an informative yet incomplete view of the factors driving predictions in S2T models.

Related papers

• Explaining multimodal LLMs via intra-modal token interactions [55.27436637894534]
  Multimodal Large Language Models (MLLMs) have achieved remarkable success across diverse vision-language tasks, yet their internal decision-making mechanisms remain insufficiently understood.<n>We propose enhancing interpretability by leveraging intra-modal interaction.
  arXiv  Detail & Related papers  (2025-09-26T14:39:13Z)
• Entropy-based Coarse and Compressed Semantic Speech Representation Learning [72.18542411704347]
  We propose an entropy-based dynamic aggregation framework for learning compressed semantic speech representations.<n> Experiments on ASR, speech-to-text translation, and voice conversion tasks demonstrate that the compressed representations perform on par with or better than dense token sequences.
  arXiv  Detail & Related papers  (2025-08-30T13:50:58Z)
• T2ICount: Enhancing Cross-modal Understanding for Zero-Shot Counting [20.21019748095159]
  Zero-shot object counting aims to count instances of arbitrary object categories specified by text descriptions.<n>We present T2ICount, a diffusion-based framework that leverages rich prior knowledge and fine-grained visual understanding from pretrained diffusion models.
  arXiv  Detail & Related papers  (2025-02-28T01:09:18Z)
• Weight-based Analysis of Detokenization in Language Models: Understanding the First Stage of Inference Without Inference [30.31106907785379]
  We show that several important aspects of the detokenization stage can be understood purely by analyzing model weights.<n>Our decomposition yields interpretable terms that quantify the relative contributions of position-related, token-related, and mixed effects.
  arXiv  Detail & Related papers  (2025-01-27T03:45:29Z)
• Core Context Aware Transformers for Long Context Language Modeling [50.774702091154204]
  We propose a plug-and-play Core Context Aware (CCA) Attention for efficient long-context modeling.<n>Our method automatically focuses and strengthens core context while diminishing redundancy during the learning process.<n>Our method is able to replace the self-attention module in existing Large Language Models with minimal fine-tuning cost.
  arXiv  Detail & Related papers  (2024-12-17T01:54:08Z)
• Coreference-aware Double-channel Attention Network for Multi-party Dialogue Reading Comprehension [7.353227696624305]
  We tackle Multi-party Dialogue Reading (abbr., MDRC) MDRC stands for an extractive reading comprehension task grounded on a batch of dialogues among multiple interlocutors. We propose a coreference-aware attention modeling method to strengthen the reasoning ability.
  arXiv  Detail & Related papers  (2023-05-15T05:01:29Z)
• Leveraging Acoustic Contextual Representation by Audio-textual Cross-modal Learning for Conversational ASR [25.75615870266786]
  We propose an audio-textual cross-modal representation extractor to learn contextual representations directly from preceding speech. The effectiveness of the proposed approach is validated on several Mandarin conversation corpora.
  arXiv  Detail & Related papers  (2022-07-03T13:32:24Z)
• Analysis of Joint Speech-Text Embeddings for Semantic Matching [3.6423306784901235]
  We study a joint speech-text embedding space trained for semantic matching by minimizing the distance between paired utterance and transcription inputs. We extend our method to incorporate automatic speech recognition through both pretraining and multitask scenarios.
  arXiv  Detail & Related papers  (2022-04-04T04:50:32Z)
• Direct speech-to-speech translation with discrete units [64.19830539866072]
  We present a direct speech-to-speech translation (S2ST) model that translates speech from one language to speech in another language without relying on intermediate text generation. We propose to predict the self-supervised discrete representations learned from an unlabeled speech corpus instead. When target text transcripts are available, we design a multitask learning framework with joint speech and text training that enables the model to generate dual mode output (speech and text) simultaneously in the same inference pass.
  arXiv  Detail & Related papers  (2021-07-12T17:40:43Z)
• Question Answering Infused Pre-training of General-Purpose Contextualized Representations [70.62967781515127]
  We propose a pre-training objective based on question answering (QA) for learning general-purpose contextual representations. We accomplish this goal by training a bi-encoder QA model, which independently encodes passages and questions, to match the predictions of a more accurate cross-encoder model. We show large improvements over both RoBERTa-large and previous state-of-the-art results on zero-shot and few-shot paraphrase detection.
  arXiv  Detail & Related papers  (2021-06-15T14:45:15Z)
• SPLAT: Speech-Language Joint Pre-Training for Spoken Language Understanding [61.02342238771685]
  Spoken language understanding requires a model to analyze input acoustic signal to understand its linguistic content and make predictions. Various pre-training methods have been proposed to learn rich representations from large-scale unannotated speech and text. We propose a novel semi-supervised learning framework, SPLAT, to jointly pre-train the speech and language modules.
  arXiv  Detail & Related papers  (2020-10-05T19:29:49Z)
• Salience Estimation with Multi-Attention Learning for Abstractive Text Summarization [86.45110800123216]
  In the task of text summarization, salience estimation for words, phrases or sentences is a critical component. We propose a Multi-Attention Learning framework which contains two new attention learning components for salience estimation.
  arXiv  Detail & Related papers  (2020-04-07T02:38:56Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.