Research map / Vision-language and multimodal / Cross-modal learning
Reconcile Gradient Modulation for Harmony Multimodal Learning
Xiyuan Gao, Bing Cao, Baoquan Gong, Pengfei Zhu
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 all three modules (EAR, Syn, Orth) to joint training (ablation) | Joint training ResNet-18 · Kinetics Sounds | accuracy: 65.74 -> 72.56 | Improved |
| Add all three modules to joint training (ablation) | Joint training ResNet-18 · CREMA-D | accuracy: 67.47 -> 74.91 | Improved |
| Add all three modules to joint training (ablation) | Joint training ResNet-18 · UCF-51 | accuracy: 68.23 -> 76.30 | Improved |
| Each single module added alone (ablation) | Joint training · KS, CREMA-D, UCF-51 | accuracy | Improved |
| Combine all three modules rather than any single one (ablation) | RGM · KS, CREMA-D, UCF-51 | accuracy | Improved |
| RGM vs best baseline Remixing/InfoReg on Kinetics Sounds (CNN) | InfoReg · Kinetics Sounds ResNet-18 | accuracy: 72.03 -> 72.56 | Improved |
| RGM vs Diag&Re on CREMA-D (CNN) | Diag&Re · CREMA-D ResNet-18 | accuracy: 75.13 -> 74.91 | No clear effect |
| RGM vs CGGM on Kinetics Sounds ViT | CGGM · Kinetics Sounds ViT-Base | accuracy: 59.82 -> 60.19 | Improved |
| PCGrad conflict alignment on balance | Joint training · Kinetics Sounds | balance factor | Hurt |
Closest papers by meaning
- Modality-Aware SAM: Sharpness-Aware-Minimization Driven Gradient Modulation for Harmonized Multimodal Learning (NeurIPS 2025 · similarity 0.91)
- Asymmetric Reinforcing Against Multi-Modal Representation Bias (AAAI 2025 · similarity 0.90)
- Multimodal Negative Learning (NeurIPS 2025 · similarity 0.89)
- Improving Multimodal Learning Balance and Sufficiency through Data Remixing (ICML 2025 · similarity 0.89)
- Enhancing Multimodal Cooperation via Sample-level Modality Valuation (CVPR 2024 · similarity 0.89)
- Classifier-guided Gradient Modulation for Enhanced Multimodal Learning (NeurIPS 2024 · similarity 0.89)
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.
