Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Inference-Time Dynamic Modality Selection for Incomplete Multimodal Classification

About

Multimodal deep learning (MDL) has achieved remarkable success across various domains, yet its practical deployment is often hindered by incomplete multimodal data. Existing incomplete MDL methods either discard missing modalities, risking the loss of valuable task-relevant information, or recover them, potentially introducing irrelevant noise, leading to the discarding-imputation dilemma. To address this dilemma, in this paper, we propose DyMo, a new inference-time dynamic modality selection framework that adaptively identifies and integrates reliable recovered modalities, fully exploring task-relevant information beyond the conventional discard-or-impute paradigm. Central to DyMo is a novel selection algorithm that maximizes multimodal task-relevant information for each test sample. Since direct estimation of such information at test time is intractable due to the unknown data distribution, we theoretically establish a connection between information and the task loss, which we compute at inference time as a tractable proxy. Building on this, a novel principled reward function is proposed to guide modality selection. In addition, we design a flexible multimodal network architecture compatible with arbitrary modality combinations, alongside a tailored training strategy for robust representation learning. Extensive experiments on diverse natural and medical image datasets show that DyMo significantly outperforms state-of-the-art incomplete/dynamic MDL methods across various missing-data scenarios. Our code is available at https://github.com//siyi-wind/DyMo.

Siyi Du, Xinzhe Luo, Declan P. O'Regan, Chen Qin• 2026

Related benchmarks

TaskDatasetResultRank
Multimodal ClassificationMST Missing Modalities
Accuracy100
28
Multimodal ClassificationPolyMNIST Missing Rate η=0.6
Accuracy99.87
16
Multimodal ClassificationPolyMNIST Missing Rate η=0.8
Accuracy96.81
16
Multimodal ClassificationPolyMNIST Missing Rate η=0
Accuracy100
16
Multimodal ClassificationCelebA Missing Modalities {I}
Accuracy100
14
Multimodal ClassificationCelebA Missing Modalities {T}
Accuracy95.2
14
Multimodal ClassificationMST Missing Modalities {S,T}
Accuracy0.9843
14
Multimodal ClassificationCelebA Missing Modalities
Accuracy100
14
Multimodal ClassificationDVM Missing Tabular Rate γ=0
Accuracy99.5
10
Multimodal ClassificationCAD Missing Tabular Rate γ=0
AUC86.17
10
Showing 10 of 17 rows

Other info

Follow for update