Saved in:
Bibliographic Details
Main Authors: Magal, Nicholas, Tran, Minh, Arakawa, Riku, Nie, Suzanne
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2501.00865
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913632620642304
author Magal, Nicholas
Tran, Minh
Arakawa, Riku
Nie, Suzanne
author_facet Magal, Nicholas
Tran, Minh
Arakawa, Riku
Nie, Suzanne
contents This paper aims to document an effective way to improve multimodal co-learning by using aggressive modality dropout. We find that by using aggressive modality dropout we are able to reverse negative co-learning (NCL) to positive co-learning (PCL). Aggressive modality dropout can be used to "prep" a multimodal model for unimodal deployment, and dramatically increases model performance during negative co-learning, where during some experiments we saw a 20% gain in accuracy. We also benchmark our modality dropout technique against PCL to show that our modality drop out technique improves co-learning during PCL, although it does not have as much as an substantial effect as it does during NCL. Github: https://github.com/nmagal/modality_drop_for_colearning
format Preprint
id arxiv_https___arxiv_org_abs_2501_00865
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Negative to Positive Co-learning with Aggressive Modality Dropout
Magal, Nicholas
Tran, Minh
Arakawa, Riku
Nie, Suzanne
Computation and Language
Machine Learning
This paper aims to document an effective way to improve multimodal co-learning by using aggressive modality dropout. We find that by using aggressive modality dropout we are able to reverse negative co-learning (NCL) to positive co-learning (PCL). Aggressive modality dropout can be used to "prep" a multimodal model for unimodal deployment, and dramatically increases model performance during negative co-learning, where during some experiments we saw a 20% gain in accuracy. We also benchmark our modality dropout technique against PCL to show that our modality drop out technique improves co-learning during PCL, although it does not have as much as an substantial effect as it does during NCL. Github: https://github.com/nmagal/modality_drop_for_colearning
title Negative to Positive Co-learning with Aggressive Modality Dropout
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2501.00865