AI Paper Atlas

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

SeD-UD: An Influence-Driven and Hierarchically-Decoupled Information Bottleneck for Multimodal Intent Recognition

Qin Li, Wenbo Zhang, Limei Liu, Han Peng, Junfeng Yang, Guanying Xu

CVPR 2026 · Vision-language and multimodal › Cross-modal learning · approach: Multimodal sentiment analysis

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
replace fixed-dimension IB with input-adaptive IDAB (ablation)SeD-UD · MIntRecACC (%): 71.47 -> 73.81Improved
use dynamic compression dimension instead of fixed average dimension (ablation)IDAB with ranking selection · MIntRecACC (%): 69.99 -> 73.81Improved
use importance-ranking parameter selection instead of random (ablation)IDAB with dynamic Dc · MIntRecACC (%): 71.91 -> 73.81Improved
add denoising on fused feature (ablation)SeD-UD · MIntRecACC (%): 68.54 -> 71.69Improved
add unimodal de-redundancy (ablation)SeD-UD · MIntRecACC (%): 68.54 -> 69.21Improved
combine de-redundancy on all three modalities with denoising (ablation)SeD-UD · MELD-DAACC (%): 62.66 -> 63.72Improved
add text de-redundancy (ablation)SeD-UD · MIntRecACC (%): 71.97 -> 73.81Improved
denoise after fusion (DR->MF->MD) instead of per-modality denoising (ablation)SeD-UD · MIntRecACC (%): 71.24 -> 73.81Improved
perform de-redundancy before denoising (ablation)SeD-UD · MIntRecACC (%): 70.63 -> 71.24Improved
use modality modulation-based fusion instead of linear mapping (ablation)SeD-UD · MIntRecACC (%): 64.33 -> 73.81Improved
use modality modulation fusion instead of MAP (ablation)SeD-UD · MIntRecACC (%): 72.93 -> 73.81Improved
set Dmin to 64 instead of 32 or 128 (ablation)IDAB · MIntRecACC (%): 70.56 -> 73.81Improved

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