Research map / Vision-language and multimodal / Cross-modal learning
SimMMDG: A Simple and Effective Framework for Multi-modal Domain Generalization
Hao Dong, Ismail Nejjar, Han Sun, Eleni Chatzi, Olga Fink
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 |
|---|---|---|---|
| Use SimMMDG with video, audio, and optical flow instead of the strongest listed baseline | Multi-source DG · EPIC-Kitchens | Top-1 accuracy: 56.59 -> 67.19 | Improved |
| Use SimMMDG with video, audio, and optical flow instead of the strongest listed baseline | Multi-source DG · HAC | Top-1 accuracy: 66.74 -> 71.10 | Improved |
| Replace zero-filled missing audio features with cross-modal translated features | SimMMDG · EPIC-Kitchens, video available and audio missing | Top-1 accuracy: 59.24 -> 60.65 | Improved |
| Replace zero-filled missing video features with cross-modal translated features | SimMMDG · EPIC-Kitchens, audio available and video missing | Top-1 accuracy: 37.41 -> 45.38 | Improved |
| Add supervised contrastive learning (ablation) | SimMMDG · EPIC-Kitchens multi-source DG | Top-1 accuracy: 55.45 -> 58.67 | Improved |
| Add feature splitting without constraints (ablation) | SimMMDG with supervised contrastive learning · EPIC-Kitchens multi-source DG | Top-1 accuracy: 58.67 -> 57.77 | Hurt |
| Add distance loss to promote diversity between modality-specific and modality-shared features (ablation) | SimMMDG with supervised contrastive learning and feature splitting · EPIC-Kitchens multi-source DG | Top-1 accuracy: 57.77 -> 61.18 | Improved |
| Add cross-modal translation (ablation) | SimMMDG with supervised contrastive learning, feature splitting, and distance loss · EPIC-Kitchens multi-source DG | Top-1 accuracy: 61.18 -> 63.24 | Improved |
| Add supervised contrastive learning (ablation) | SimMMDG with feature splitting, distance loss, and cross-modal translation · EPIC-Kitchens multi-source DG | Top-1 accuracy: 60.25 -> 63.24 | Improved |
| Use ℓ2 distance rather than cosine similarity in the distance loss (ablation) | SimMMDG · EPIC-Kitchens multi-source DG | Top-1 accuracy: 58.00 -> 63.24 | Improved |
| Use ℓ2 distance rather than ℓ1 norm in the distance loss (ablation) | SimMMDG · EPIC-Kitchens multi-source DG | Top-1 accuracy: 58.95 -> 63.24 | Improved |
| Add Gradient Blending | SimMMDG · EPIC-Kitchens multi-source DG | Top-1 accuracy: 63.24 -> 64.78 | Improved |
Later papers that compared with this method
3 later accepted papers report SimMMDG as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Modality-Balanced Collaborative Distillation for Multi-Modal Domain Generalization (AAAI 2026 · better on 7 of 7 shared results)
- Bridging Domain Generalization to Multimodal Domain Generalization via Unified Representations (ICCV 2025 · better on 3 of 3 shared results)
- Cross-modal Representation Flattening for Multi-modal Domain Generalization (NeurIPS 2024 · better on 7 of 7 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
- Bridging Domain Generalization to Multimodal Domain Generalization via Unified Representations (ICCV 2025 · similarity 0.95)
- Cross-modal Representation Flattening for Multi-modal Domain Generalization (NeurIPS 2024 · similarity 0.93)
- Are We Making Progress in Multimodal Domain Generalization? A Comprehensive Benchmark Study (NeurIPS 2026 · similarity 0.89)
- Multimodal Unsupervised Domain Generalization by Retrieving Across the Modality Gap (ICLR 2025 · similarity 0.88)
- Connect, Collapse, Corrupt: Learning Cross-Modal Tasks with Uni-Modal Data (ICLR 2024 · similarity 0.88)
- Jointly Modeling Inter- & Intra-Modality Dependencies for Multi-modal Learning (NeurIPS 2024 · 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.
