Research map / Vision-language and multimodal / Cross-modal learning
BALM: A Model-Agnostic Framework for Balanced Multimodal Learning under Imbalanced Missing Rates
Phuong-Anh Nguyen, Tien Anh Pham, Duc-Trong Le, Cam-Van Thi Nguyen
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 BALM plug-in | MMDFN · IEMOCAP IMR (0.3,0.5,0.7) | accuracy: 63.34 -> 67.34 | Improved |
| add BALM plug-in | MMGCN · IEMOCAP IMR (0.5,0.7,0.3) | accuracy: 57.79 -> 62.29 | Improved |
| add BALM plug-in | GCNet · IEMOCAP IMR settings | accuracy: +3.27 to -0.31 | Mixed |
| add BALM plug-in | MMGCN · CMU-MOSEI IMR settings | accuracy: +1.71 to -0.22 | Mixed |
| add BALM plug-in | MMDFN · CMU-MOSEI IMR (0.3,0.5,0.7) | accuracy: 82.17 -> 82.06 | No clear effect |
| add Feature Calibration Module (ablation) | MMGCN · IEMOCAP average | w-F1: 44.81 -> 47.62 | Improved |
| add Feature Calibration Module (ablation) | MMGCN · CMU-MOSEI average | w-F1: 69.17 -> 70.09 | Improved |
| add Gradient Rebalancing Module after FCM (ablation) | MMGCN · IEMOCAP | cosine similarity and KL divergence | Improved |
| add BALM under 90% visual missing | MMGCN · IEMOCAP severe IMR (0.1,0.1,0.9) | accuracy: 68.02 -> 65.87 | Hurt |
| add BALM under 90% language missing | MMGCN · IEMOCAP severe IMR (0.1,0.9,0.1) | accuracy: 52.74 -> 58.72 | Improved |
| vary hyperparameters tau and rho (ablation) | MMGCN+BALM · IEMOCAP Config-A (0.5,0.3,0.7) | w-F1 | No clear effect |
Closest papers by meaning
- A Multi-Focus-Driven Multi-Branch Network for Robust Multimodal Sentiment Analysis (AAAI 2025 · similarity 0.87)
- RedCore: Relative Advantage Aware Cross-Modal Representation Learning for Missing Modalities with Imbalanced Missing Rates (AAAI 2024 · similarity 0.87)
- MIDAS: Misalignment-based Data Augmentation Strategy for Imbalanced Multimodal Learning (NeurIPS 2025 · similarity 0.87)
- Multi-to-Single: Reducing Multimodal Dependency in Emotion Recognition Through Contrastive Learning (AAAI 2025 · similarity 0.87)
- Asymmetric Reinforcing Against Multi-Modal Representation Bias (AAAI 2025 · similarity 0.87)
- Rethinking Multimodal Learning from the Perspective of Mitigating Classification Ability Disproportion (NeurIPS 2025 · similarity 0.86)
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.
