SeD-UD：面向多模态意图识别的、影响驱动且分层解耦的信息瓶颈
SeD-UD: An Influence-Driven and Hierarchically-Decoupled Information Bottleneck for Multimodal Intent Recognition
论文提出的影响驱动自适应瓶颈（IDAB）模块可根据样本级冗余/噪声因子动态调整瓶颈维度，这一思路可直接迁移到视觉-语言模型的多模态特征融合环节，用于按样本抑制视觉与文本模态中的冗余信息。
原文依据 IDAB dynamically adjusts dimensions and selects the optimal parameters for compression and reconstruction
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论多模态意图识别（MIR）受到来自文本、语音和视觉输入的大量冗余与噪声的阻碍，这些冗余与噪声削弱了特征的判别性，最终损害识别性能。尽管近期基于信息瓶颈（IB）原理的方法通过特征压缩与重建来获得紧凑且降噪的表示，从而缓解了该问题，但它们仍面临两大缺陷。第一，传统 IB 采用固定的瓶颈维度，无法适应随样本而变的冗余与噪声。第二，在单一压缩过程中同时处理冗余与噪声会导致特征净化不彻底。本文提出一种名为 SeD-UD 的新框架，它在分层解耦的 IB 结构之后引入影响驱动的输入自适应瓶颈（IDAB）模块。给定冗余/噪声影响因子，IDAB 动态调整维度并选择最优参数进行压缩与重建，从而在信息保留与干扰抑制之间取得最佳折衷。该 IB 结构通过基于 IDAB 模块的分离式去冗余与统一去噪，对冗余和噪声进行分层解耦处理。在基准数据集上的大量实验表明，SeD-UD 优于当前最优模型。
英文原文摘要 1426 字符
Multimodal intent recognition (MIR) is hindered by substantial redundancy and noise originating from text, speech, and visual inputs, which weakens feature distinctiveness and ultimately harms recognition performance. Although recent approaches based on the information bottleneck (IB) principle mitigate this issue via feature compression and reconstruction to obtain compact and noise-reduced representations, they still encounter two major drawbacks. First, conventional IB employs a fixed bottleneck dimension, making it unable to accommodate sample-dependent variations in redundancy and noise. Second, simultaneously handling redundancy and noise within a single compression process leads to incomplete feature purification. In this paper, we propose a novel framework named SeD-UD, which incorporates influence-driven input-adaptive bottleneck (IDAB) modules following a hierarchically-decoupled IB structure. Given a redundancy/noise influence factor, IDAB dynamically adjusts dimensions and selects the optimal parameters for compression and reconstruction, thereby achieving the best trade-off between information preservation and interference suppression. The IB structure performs hierarchically-decoupled processing of redundancy and noise via separated de-redundancy and unified denoising based on IDAB modules. Extensive experiments on benchmark datasets show SeD-UD outperforms current state-of-the-art models.