AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

Learning to Rebalance Multi-Modal Optimization by Adaptively Masking Subnetworks

Yang Yang, Hongpeng Pan, Qing-Yuan Jiang, Yi Xu, Jinhui Tang

TPAMI 2025 · Vision-language and multimodal › Cross-modal learning · approach: Medical multimodal fusion · 23 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add AMSS+ rebalancing over concatenation fusionConcat · Kinetics-Soundaccuracy: 64.55 -> 72.25Improved
add AMSS+ rebalancing over concatenation fusionConcat · CREMA-Daccuracy: 63.31 -> 70.30Improved
use unbiased AMSS+ mask instead of biased AMSSAMSS · Kinetics-Soundaccuracy: 68.96 -> 72.25Improved
use unbiased AMSS+ mask instead of biased AMSSAMSS · Twitter-15accuracy: 75.89 -> 75.12Hurt
use non-uniform adaptive sampling instead of uniform random sampling (ablation)AMSS+ · Kinetics-Soundaccuracy: 66.45 -> 72.25Improved
use non-uniform adaptive sampling instead of uniform random sampling (ablation)AMSS+ · CREMA-Daccuracy: 65.59 -> 70.16Improved
add backbone mask in addition to classifier mask (ablation)AMSS · Kinetics-Soundaccuracy: 65.37 -> 68.96Improved
add classifier mask in addition to backbone mask (ablation)AMSS+ · Kinetics-Soundaccuracy: 69.15 -> 72.25Improved
set tau below 1 (0.25) instead of 1 (ablation)AMSS+ · Kinetics-Soundaccuracy: 70.20 -> 72.25Improved
set tau very small (0.10) instead of 0.20 (ablation)AMSS+ · Kinetics-Soundaccuracy: 72.52 -> 72.17Hurt
scale with 1/(p_j+L) without masking (rho=1) (ablation)AMSS+ · Kinetics-Soundaccuracy: 72.25 -> 69.39Hurt
apply AMSS+ to Affine fusionAffine · Kinetics-Soundaccuracy: 64.24 -> 69.08Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy