Research map / Vision-language and multimodal / Cross-modal learning
Self-Balancing Multimodal Models via Multi-Loss Gradient Modulation
Konstantinos Kontras, Christos Chatzichristos, Matthew B. Blaschko, Maarten De Vos
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 unimodal cross-entropy losses (Multi-Loss) to joint training (ablation) | multimodal audio-video models · CREMA-D, AVE, UCF101 | accuracy | Improved |
| add adaptive bidirectional balancing on top of Multi-Loss (ablation) | Multi-Loss · CREMA-D, AVE, UCF101 | accuracy | Improved |
| add balancing to Multi-Loss with FiLM, Gated and TF fusion (ablation) | Multi-Loss · CREMA-D (FiLM, Gated, TF fusion) | accuracy | Mixed |
| use MLB instead of Multi-Loss (ablation) | Conformer · CREMA-D | expected calibration error | Improved |
Closest papers by meaning
- Improving Multimodal Learning Balance and Sufficiency through Data Remixing (ICML 2025 · similarity 0.91)
- Modality-Aware SAM: Sharpness-Aware-Minimization Driven Gradient Modulation for Harmonized Multimodal Learning (NeurIPS 2025 · similarity 0.90)
- Asymmetric Reinforcing Against Multi-Modal Representation Bias (AAAI 2025 · similarity 0.90)
- Fusion or Confusion? Multimodal Complexity Is Not All You Need (NeurIPS 2026 · similarity 0.89)
- MSP: Modality Self-Play for Training Multimodal LLMs Without Paired Cross-Modal Data (NeurIPS 2026 · similarity 0.89)
- Mirasol3B: A Multimodal Autoregressive Model for Time-Aligned and Contextual Modalities (CVPR 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.
