AI Paper Atlas

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

THE MORE, THE MERRIER: CONTRASTIVE FUSION FOR HIGHER-ORDER MULTIMODAL ALIGNMENT

Stefanos Koutoupis, Michaela Areti Zervou, Konstantinos Kontras, Maarten De Vos, Panagiotis Tsakalides, Grigorios Tsagkatakis

CVPR 2026 · Vision-language and multimodal › Cross-modal learning · approach: Multimodal fusion with missing modalities · 2 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 fused-modality (2->1) contrastive term to pairwise objectiveConFu · AV-MNIST zero-shotaccuracy: 70.9 -> 71.2 vs SymileImproved
fuse audio and vision instead of using the strongest unimodal modelConFu · AV-MNIST zero-shotaccuracy: 8% improvement over strongest unimodal baselineImproved
fuse audio with weak-audio visual inputs on VB100ConFu · VB100 zero-shot multi-frameaccuracy: 20.69 -> 18.15 (vs best visual-only)Hurt
add audio fusion when both modalities are informativeConFu · SSW60 zero-shot multi-frameaccuracy: 70.1 -> 71.4Improved
add feature-level modality masking (ratio 0.3) (ablation)ConFu · MOSI retrievalRecall@10: +0.93 to +7.29 across directionsImproved
add feature-level modality masking (ablation)ConFu · MUStARDaccuracy and retrievalHurt
add feature-level modality masking (ablation)ConFu · UR-FUNNYaccuracy and retrievalNo clear effect
increase lambda from 0.5 to 0.8 (ablation)ConFu · MOSI mean Recall@10 1->1mean Recall@10: 19.5 -> 13.5Hurt
lambda 0.5 instead of 0.2 or 0.8 (ablation)ConFu · MUStARDaccuracy: 59.7 / 61.1 -> 64.1Improved
use ConFu instead of pairwise-only Tri-CLIP under vision noiseConFu · Vision noise 20dB SNRaccuracy: 35.4 -> 45.4Improved
embedding dimension 64 needed for ConFu on XOR vs 8 for Symile (ablation)ConFu · XOR task (p=1)accuracyMixed
add third modality fusion to pairwise CLIPConFu · synthetic XORaccuracy: Tri-CLIP ~3% (chance) vs ConFu above chanceImproved

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