Research map / Vision-language and multimodal / Cross-modal learning
Multimodal Classification via Total Correlation Maximization
Feng Yu, Xiangyu Wu, Yang Yang, Jianfeng Lu
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 |
|---|---|---|---|
| replace joint cross-entropy with TCMax total-correlation loss (Concat head) | Concat multimodal ResNet-18 · CREMA-D | multimodal accuracy: 68.5 -> 82.8 | Improved |
| replace joint cross-entropy with TCMax loss (Concat head) | Concat multimodal ResNet-18 · VGGSound | multimodal accuracy: 41.1 -> 47.6 | Improved |
| TCMax versus unimodal ensemble | multimodal ResNet-18 · Kinetics-Sounds | multimodal accuracy: 62.3 -> 62.4 (Concat); 62.3 -> 63.5 (Share Head) | No clear effect |
| TCMax versus QMF | multimodal ResNet-18 · AVE | multimodal accuracy: 65.1 -> 64.5 | Hurt |
| TCMax versus MMPareto | multimodal ResNet-18 · UCF101 | multimodal accuracy: 55.9 -> 56.0 | No clear effect |
| use TCMax instead of joint loss for modality balance | multimodal ResNet-18 · CREMA-D | visual-only accuracy: 26.5 -> 75.0 | Improved |
| use TCMax loss | multimodal ResNet-18 · CREMA-D / all datasets | JS divergence between modality predictions: 0.478 -> 0.271 (CREMA-D, Concat vs TCMax Share Head) | Improved |
| use TCMax with frozen CLIP encoders | MVSA linear classifier · MVSA (ViT-B/32) | multimodal accuracy: 82.83 -> 84.05 | Improved |
| use unimodal loss with frozen CLIP encoders | MVSA linear classifier · MVSA (RN50) | multimodal accuracy: 81.23 -> 80.02 | Hurt |
| increase number of sampled negative pairs (ablation) | TCMax · CREMA-D / UCF101 | accuracy | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MMPareto — MMPareto: Boosting Multimodal Learning with Innocent Unimodal Assistance (ICML 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Beyond Modality Fusion: Deep Ensembles for Multimodal Classification (NeurIPS 2026 · similarity 0.89)
- Asymmetric Reinforcing Against Multi-Modal Representation Bias (AAAI 2025 · similarity 0.88)
- Semi-Supervised Multimodal Classification Through Learning from Modal and Strategic Complementarities (AAAI 2025 · similarity 0.88)
- Multimodal Representation Learning by Alternating Unimodal Adaptation (CVPR 2024 · similarity 0.88)
- Rethinking Multimodal Learning from the Perspective of Mitigating Classification Ability Disproportion (NeurIPS 2025 · similarity 0.88)
- Multimodal Learning Without Labeled Multimodal Data: Guarantees and Applications (ICLR 2024 · 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.
