Research map / Vision-language and multimodal / Cross-modal learning
Back-Modality: Leveraging Modal Transformation for Data Augmentation
Zhi Li, Yifan Liu, Yin Zhang
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 multi-captioning (ablation) | Back-captioning · Tiny ImageNet, 10-shot | top-1 accuracy: 17.21 -> 20.07 | Improved |
| add GPT caption augmentation (ablation) | Back-captioning · Tiny ImageNet, 10-shot | top-1 accuracy: 18.49 -> 20.07 | Improved |
| add multi-imagination (ablation) | Back-captioning · Tiny ImageNet, 10-shot | top-1 accuracy: 18.92 -> 20.07 | Improved |
| add multi-imagination (ablation) | Back-imagination · TNCC, 5-shot | accuracy: 84.71 -> 88.41 | Improved |
| add multi-captioning (ablation) | Back-imagination · TNCC, 5-shot | accuracy: 86.52 -> 88.41 | Improved |
| add pitch shifting (ablation) | Back-speech · SST-2, 5-shot | accuracy: 58.45 -> 59.03 | Improved |
| add time stretching (ablation) | Back-speech · SST-2, 5-shot | accuracy: 58.60 -> 59.03 | Improved |
| augment with back-captioning vs no augmentation | ResNet-18 · Tiny ImageNet, 10-shot | top-1 accuracy: 11.75 -> 20.07 | Improved |
| use back-captioning instead of Puzzle Mix | ResNet-18 · Tiny ImageNet, 10-shot | top-1 accuracy: 15.66 -> 20.07 | Improved |
| augment with back-imagination vs Treemix | BERT-base · TNCC, 1-shot | accuracy: 62.88 -> 69.20 | Improved |
| augment with back-speech vs no augmentation | BERT-base · SST-2, 1-shot to 10-shot | accuracy: 51.46 -> 52.13 (1-shot); 61.10 -> 63.21 (10-shot) | Improved |
| apply EDA augmentation at 1-shot | BERT-base · SST-2, 1-shot | accuracy: 51.46 -> 50.35 | Hurt |
Closest papers by meaning
- Strong Helps Weak: Directional Cross-Modal Alignment Transfer in Multi-modal LLMs (NeurIPS 2026 · similarity 0.86)
- Learning Modality Knowledge Alignment for Cross-Modality Transfer (ICML 2024 · similarity 0.85)
- MDReID: Modality-Decoupled Learning for Any-to-Any Multi-Modal Object Re-Identification (NeurIPS 2025 · similarity 0.85)
- Cross the Gap: Exposing the Intra-modal Misalignment in CLIP via Modality Inversion (ICLR 2025 · similarity 0.85)
- ReconBoost: Boosting Can Achieve Modality Reconcilement (ICML 2024 · similarity 0.84)
- ModaVerse: Efficiently Transforming Modalities with LLMs (CVPR 2024 · similarity 0.84)
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.
