Research map / Vision-language and multimodal / Cross-modal learning
Plug, Play, and Fortify: A Low-Cost Module for Robust Multimodal Image Understanding Models
Siqi Lu, Wanying XU, Yongbin Zheng, Wenting Luan, Peng Sun, Jianhang Yao
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add MWAM | RFNet · BRATS2020 | average Dice: 85.41 -> 86.41 | Improved |
| add MWAM | GSS · BRATS2020 | average Dice: 86.41 -> 87.56 | Improved |
| add MWAM | mmFormer (ViT-based) · BRATS2020 | average Dice: 85.64 -> 86.30 | Improved |
| add MWAM | ESANet-MD · NYU-Depth V2 | average MIoU: 42.53 -> 44.47 | Improved |
| add MWAM | MMANet · NYU-Depth V2 | average MIoU: 43.47 -> 45.81 | Improved |
| add MWAM | SF-MD · CASIA-SURF | average accuracy: 92.85 -> 96.41 | Improved |
| add MWAM | MMANet · CASIA-SURF | average accuracy: 96.06 -> 97.03 | Improved |
| add MWAM | MMANet · CASIA-SURF, full RGB+Depth+IR input | accuracy: 99.13 -> 98.74 | Hurt |
| add MWAM | RFNet · BRATS2020, Flair+T1+T1c | Dice: 89.36 -> 89.84 Dice; PCR 0.87 -> 1.01 | Mixed |
| loss-level intervention only (vs none) (ablation) | MWAM on SF-MD · CASIA-SURF average | accuracy: 92.85 -> 95.06 | Improved |
| gradient-level instead of loss-level intervention (ablation) | MWAM on SF-MD · CASIA-SURF average | accuracy: 95.06 -> 95.73 | Improved |
| hybrid loss+gradient instead of gradient only (ablation) | MWAM on SF-MD · CASIA-SURF average | accuracy: 95.73 -> 96.41 | Improved |
Closest papers by meaning
- Self-Balancing Multimodal Models via Multi-Loss Gradient Modulation (IJCV 2026 · similarity 0.90)
- Modality-Aware SAM: Sharpness-Aware-Minimization Driven Gradient Modulation for Harmonized Multimodal Learning (NeurIPS 2025 · similarity 0.89)
- Deep Correlated Prompting for Visual Recognition with Missing Modalities (NeurIPS 2024 · similarity 0.89)
- Do More Modalities Always Help? A Geometric Perspective on Missing-Modality Robustness (NeurIPS 2026 · similarity 0.89)
- UNIALIGN: Scaling Multimodal Alignment within One Unified Model (CVPR 2025 · similarity 0.88)
- Asymmetric Reinforcing Against Multi-Modal Representation Bias (AAAI 2025 · similarity 0.88)
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.
