Research map / Vision-language and multimodal / Cross-modal learning
MASAM: Multimodal Adaptive Sharpness-Aware Minimization for Heterogeneous Data Fusion
Zijie Chen, Kejing Yin, Wenfang Yao, William K. Cheung, Jing Qin
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 APS-based dominant-modality selection (ablation) | MASAM · MIMIC-IV Phenotype | AUPRC: 0.484 -> 0.498 | Improved |
| add APS-based dominant-modality selection (ablation) | MASAM · Kinetics-Sounds | accuracy: 0.724 -> 0.740 | Improved |
| add MDPS decoupled perturbation scaling (ablation) | MASAM · MIMIC-IV Phenotype | AUPRC: 0.491 -> 0.498 | Improved |
| add MDPS decoupled perturbation scaling (ablation) | MASAM · Kinetics-Sounds | accuracy: 0.723 -> 0.740 | Improved |
| add APS and MDPS to plain SAM (ablation) | SAM · Kinetics-Sounds | accuracy: 0.689 -> 0.740 | Improved |
| add plain SAM to late fusion (ablation) | Late Fusion · MIMIC-IV Phenotype | AUPRC: 0.475 -> 0.478 | Improved |
| add plain SAM to late fusion (ablation) | Late Fusion · Kinetics-Sounds | accuracy: 0.636 -> 0.689 | Improved |
| replace Late Fusion with MASAM | Late Fusion · CREMA-D | accuracy: 0.660 -> 0.814 | Improved |
| replace Late Fusion with MASAM | Late Fusion · UPMC-Food101 | accuracy: 0.907 -> 0.935 | Improved |
| replace best gradient-modulation baseline with MASAM | Modality-imbalance baselines · ADNI | mAP: 0.847 -> 0.857 | Improved |
| replace Late Fusion with MASAM | Late Fusion · CREMA-D (60% label noise) | accuracy: 0.389 -> 0.460 | Improved |
| replace Late Fusion with MASAM | Late Fusion · Kinetics-Sounds (60% label noise) | accuracy: 0.387 -> 0.501 | Improved |
Closest papers by meaning
- Modality-Aware SAM: Sharpness-Aware-Minimization Driven Gradient Modulation for Harmonized Multimodal Learning (NeurIPS 2025 · similarity 0.93)
- Understanding Multimodal Learning From Modality Fusion and Alignment Perspectives (TPAMI 2026 · similarity 0.90)
- Self-Balancing Multimodal Models via Multi-Loss Gradient Modulation (IJCV 2026 · similarity 0.90)
- Asymmetric Reinforcing Against Multi-Modal Representation Bias (AAAI 2025 · similarity 0.89)
- Beyond Modality Fusion: Deep Ensembles for Multimodal Classification (NeurIPS 2026 · similarity 0.89)
- Improving Multimodal Learning Balance and Sufficiency through Data Remixing (ICML 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.
