Research map / Vision-language and multimodal / Cross-modal learning
IBMA: Information Bottleneck-Based Multimodal Alignment
Yancheng Wang, Zeyu Dong, Dongfang Sun, Alvin C Silva, Teresa Wu, Yingzhen Yang
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 modality-specific representation alignment (ablation) | IBMA · CheXpert | mAUC (%): 89.7 -> 91.1 | Improved |
| add modality-specific representation alignment (ablation) | IBMA · MIMIC-CXR | mAUC (%): 71.4 -> 72.7 | Improved |
| use IBB instead of CLUB bound (ablation) | IBMA · CheXpert | mAUC (%): 90.0 -> 91.1 | Improved |
| use IBB instead of VIB bound (ablation) | IBMA · CREMA-D | accuracy: 63.9 -> 65.4 | Improved |
| use IBB instead of APIB (ablation) | IBMA · CREMA-D | accuracy: 63.8 -> 65.4 | Improved |
| use IBB instead of CLUB for training cost (ablation) | IBMA · CREMA-D | training time (min/epoch): 3.8 -> 2.0 | Improved |
| add modality-specific IB alignment lowers IB loss (ablation) | IBMA · MIMIC-CXR | multimodal IB loss: -0.035 -> -0.058 | Improved |
| use IBMA instead of OMIB | IBMA · CREMA-D | accuracy: 63.6 -> 65.4 | Improved |
| use IBB instead of VIB (ablation) | IBMA · CREMA-D | training time (min/epoch): 1.9 -> 2.0 | Hurt |
Closest papers by meaning
- Aligning Multimodal Representations through an Information Bottleneck (ICML 2025 · similarity 0.92)
- Learning Optimal Multimodal Information Bottleneck Representations (ICML 2025 · similarity 0.89)
- On the Value of Cross-Modal Misalignment in Multimodal Representation Learning (NeurIPS 2025 · similarity 0.89)
- CyIN: Cyclic Informative Latent Space for Bridging Complete and Incomplete Multimodal Learning (NeurIPS 2025 · similarity 0.88)
- Asymmetric Reinforcing Against Multi-Modal Representation Bias (AAAI 2025 · similarity 0.88)
- With Limited Data for Multimodal Alignment, Let the STRUCTURE Guide You (NeurIPS 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.
