FANoise: Singular Value-Adaptive Noise Modulation for Robust Multimodal Representation Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Jiaoyang, Fang, Jun, Gao, Tianhao, Zhang, Xiaohui, Liu, Zhiyuan, Liu, Chao, Liu, Pengzhang, Jiang, Qixia
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917105439342592
author Li, Jiaoyang
Fang, Jun
Gao, Tianhao
Zhang, Xiaohui
Liu, Zhiyuan
Liu, Chao
Liu, Pengzhang
Jiang, Qixia
author_facet Li, Jiaoyang
Fang, Jun
Gao, Tianhao
Zhang, Xiaohui
Liu, Zhiyuan
Liu, Chao
Liu, Pengzhang
Jiang, Qixia
contents Representation learning is fundamental to modern machine learning, powering applications such as text retrieval and multimodal understanding. However, learning robust and generalizable representations remains challenging. While prior work has demonstrated that active noise injection, a form of data augmentation, can enhance encoding performance, most existing methods rely on heuristic or static noise, overlooking the dynamic nature of feature distributions during training. In this work, we systematically study the role of noise in representation learning from both gradient-based and feature distribution perspectives, using InfoNCE loss as a representative example. Focusing on multimodal representation learning, we propose FANoise, a novel feature-adaptive noise injection strategy. By leveraging the dynamics of contrastive learning, FANoise effectively mitigates the negative impacts of noise while preserving its benefits. Under this theoretically grounded framework, comprehensive experiments demonstrate that FANoise consistently improves overall performance on multimodal tasks across various base VLM models.
format Preprint
id arxiv_https___arxiv_org_abs_2511_20997
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle FANoise: Singular Value-Adaptive Noise Modulation for Robust Multimodal Representation Learning
Li, Jiaoyang
Fang, Jun
Gao, Tianhao
Zhang, Xiaohui
Liu, Zhiyuan
Liu, Chao
Liu, Pengzhang
Jiang, Qixia
Machine Learning
Artificial Intelligence
Representation learning is fundamental to modern machine learning, powering applications such as text retrieval and multimodal understanding. However, learning robust and generalizable representations remains challenging. While prior work has demonstrated that active noise injection, a form of data augmentation, can enhance encoding performance, most existing methods rely on heuristic or static noise, overlooking the dynamic nature of feature distributions during training. In this work, we systematically study the role of noise in representation learning from both gradient-based and feature distribution perspectives, using InfoNCE loss as a representative example. Focusing on multimodal representation learning, we propose FANoise, a novel feature-adaptive noise injection strategy. By leveraging the dynamics of contrastive learning, FANoise effectively mitigates the negative impacts of noise while preserving its benefits. Under this theoretically grounded framework, comprehensive experiments demonstrate that FANoise consistently improves overall performance on multimodal tasks across various base VLM models.
title FANoise: Singular Value-Adaptive Noise Modulation for Robust Multimodal Representation Learning
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2511.20997