Research map / Vision-language and multimodal / Cross-modal learning
Attention Is Not Enough: Mitigating the Distribution Discrepancy in Asynchronous Multimodal Sequence Fusion
Tao Liang, Guosheng Lin, Lei Feng, Yan Zhang, Fengmao Lv
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 MMD marginal alignment to the backbone (ablation) | MICA backbone · CMU-MOSEI | Acc2: 81.9 -> 83.2 | Improved |
| add propagated element-level alignment on top of MMD alignment (ablation) | MICA · CMU-MOSEI | F1: 82.9 -> 83.3 | Improved |
| add propagation of correlations across layers to element-level alignment (ablation) | MICA · CMU-MOSEI | Acc2: 83.4 -> 83.7 | Improved |
| add propagation across layers to element-level alignment (ablation) | MICA · CMU-MOSEI | Acc7: 52.2 -> 52.4 | Improved |
| add both alignment losses to the backbone (ablation) | MICA backbone · CMU-MOSEI | Acc7: 51.1 -> 52.4 | Improved |
| use merging gate backbone instead of MulT (ablation) | MulT · CMU-MOSEI | A-distance / Corr. Acc: 1.63 -> 1.69; 80.4 -> 81.8 | Mixed |
| add alignment losses (MICA) to the backbone (ablation) | Backbone · CMU-MOSEI | A-distance: 1.69 -> 1.37 | Improved |
| add alignment losses (MICA) to the backbone (ablation) | Backbone · CMU-MOSEI | Corr. Acc: 81.8 -> 86.7 | Improved |
| replace MulT with MICA | MulT · CMU-MOSI | Acc7: 39.1 -> 40.8 | Improved |
| replace MulT with MICA | MulT · CMU-MOSEI | Acc2: 81.6 -> 83.7 | Improved |
| replace MulT with MICA | MulT · IEMOCAP Happy | F1: 81.9 -> 83.9 | Improved |
| raise trade-off alpha from 0.2 to 0.6 (ablation) | MICA · CMU-MOSEI | F1: 82.89 -> 83.17 | No clear effect |
Closest papers by meaning
- Dissecting Multimodality in VideoQA Transformer Models by Impairing Modality Fusion (ICML 2024 · similarity 0.89)
- Mirasol3B: A Multimodal Autoregressive Model for Time-Aligned and Contextual Modalities (CVPR 2024 · similarity 0.88)
- MANGO: Multimodal Attention-based Normalizing Flow Approach to Fusion Learning (NeurIPS 2025 · similarity 0.88)
- CREMA: Generalizable and Efficient Video-Language Reasoning via Multimodal Modular Fusion (ICLR 2025 · similarity 0.87)
- LongVALE: Vision-Audio-Language-Event Benchmark Towards Time-Aware Omni-Modal Perception of Long Videos (CVPR 2025 · similarity 0.87)
- MODA: MOdular Duplex Attention for Multimodal Perception, Cognition, and Emotion Understanding (ICML 2025 · similarity 0.87)
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.
