Research map / Vision-language and multimodal / Cross-modal learning
Robust Multimodal Learning With Missing Modalities via Parameter-Efficient Adaptation
Md Kaykobad Reza, Ashley Prater-Bennette, M. Salman Asif
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 |
|---|---|---|---|
| adapt with scale and shift layers for the missing-modality case | CMNeXt (Thermal only) · MFNet | mIoU (%): 35.48 -> 50.89 | Improved |
| adapt with scale and shift layers for the missing-modality case | CMNeXt (Depth only) · NYUDv2 | mIoU (%): 5.26 -> 36.72 | Improved |
| adapt with scale and shift layers for the missing-modality case | CMNeXt (RGB only) · MCubeS | mIoU (%): 42.32 -> 50.43 | Improved |
| replace modality duplication with adaptation | CMNeXt (RGB only) · NYUDv2 | mIoU (%): 46.19 -> 52.82 | Improved |
| use adapted shared model instead of dedicated per-combination models | CMNeXt (Thermal only) · MFNet | mIoU (%): 53.34 -> 50.89 | Mixed |
| add shift parameters to scale-only adaptation (ablation) | scale-and-shift adaptation · MFNet (avg) | mIoU (%): 52.00 -> 53.06 | Improved |
| add scale parameters to shift-only adaptation (ablation) | scale-and-shift adaptation · MFNet (avg) | mIoU (%): 51.77 -> 53.06 | Improved |
| use scale and shift instead of LoRA (ablation) | parameter-efficient adaptation · MFNet (avg) | mIoU (%): 50.82 -> 53.06 | Improved |
| use scale and shift instead of BitFit (ablation) | parameter-efficient adaptation · MFNet (avg) | mIoU (%): 51.73 -> 53.06 | Improved |
| use scale and shift instead of BitFit (ablation) | parameter-efficient adaptation · NYUDv2 (avg) | mIoU (%): 44.87 -> 44.77 | No clear effect |
| use scale and shift instead of normalization-layer fine-tuning (ablation) | parameter-efficient adaptation · MFNet (avg) | mIoU (%): 51.07 -> 53.06 | Improved |
| replace input prompts with scale-and-shift adaptation | ViLT · UPMC Food-101 | accuracy (%): 78.54 -> 79.83 | Improved |
Closest papers by meaning
- Missing Modality Prediction for Unpaired Multimodal Learning via Joint Embedding of Unimodal Models (ECCV 2024 · similarity 0.91)
- Robustness in Multimodal Learning under Train-Test Modality Mismatch (ICML 2023 · similarity 0.90)
- Multimodal Negative Learning (NeurIPS 2025 · similarity 0.89)
- Do More Modalities Always Help? A Geometric Perspective on Missing-Modality Robustness (NeurIPS 2026 · similarity 0.89)
- Asymmetric Reinforcing Against Multi-Modal Representation Bias (AAAI 2025 · similarity 0.89)
- Mitigating Pervasive Modality Absence Through Multimodal Generalization and Refinement (AAAI 2025 · 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.
