Research map / Label-efficient and robust learning / Recognition and long-tail
Nonlinear Mixup: Out-Of-Manifold Data Augmentation for Text Classification
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| learn the policy mapping function for label mixing instead of fixing it (ablation) | nonlinear Mixup · SST-1 (RandomTune) | accuracy (%): 45.1 -> 47.4 | Helped |
| learn the policy mapping function instead of fixing it (ablation) | nonlinear Mixup · TREC (RandomTune) | accuracy (%): 91.2 -> 92.9 | Helped |
| learn the policy mapping function instead of fixing it (ablation) | nonlinear Mixup · Subj (RandomTune) | accuracy (%): 90.9 -> 93.0 | Helped |
| learn the policy mapping function instead of fixing it (ablation) | nonlinear Mixup · MR (RandomTune) | accuracy (%): 79.5 -> 80.6 | Helped |
| vary label embedding dimension (5 to 500) (ablation) | nonlinear Mixup · SST-1 (RandomTune) | accuracy (%): 46.14 to 47.50 | No clear effect |
| replace scalar linear mixing (wordMixup) with per-dimension nonlinear mixing and learned label mixing | wordMixup · SST-2 (RandomTune) | accuracy (%): 82.8 -> 86.2 | Helped |
| replace scalar linear mixing (wordMixup) with nonlinear mixing | wordMixup · SST-1 (RandomTune) | accuracy (%): 45.2 -> 47.4 | Helped |
| replace scalar linear mixing (wordMixup) with nonlinear mixing | wordMixup · MR (RandomTune) | accuracy (%): 78.0 -> 80.6 | Helped |
| replace senMixup with nonlinear Mixup | senMixup · Subj (PretrainTune) | accuracy (%): 94.9 -> 93.9 | Hurt |
| replace CNN baseline with nonlinear Mixup | CNNsen · MR (RandomTune) | accuracy (%): 75.5 -> 80.6 | Helped |
| replace CNN baseline with nonlinear Mixup | CNNsen · SST-2 (RandomTune) | accuracy (%): 82.3 -> 86.2 | Helped |
| replace EDA with nonlinear Mixup | EDA · TREC (RandomFix) | accuracy (%): 90.6 -> 91.2 | Helped |
Closest papers by meaning
- MixUp as Locally Linear Out-of-Manifold Regularization (AAAI 2019 · similarity 0.92)
- Tailoring Mixup to Data for Calibration (ICLR 2025 · similarity 0.87)
- First-Order Manifold Data Augmentation for Regression Learning (ICML 2024 · similarity 0.86)
- Selective Mixup Helps with Distribution Shifts, But Not (Only) because of Mixup (ICML 2024 · similarity 0.86)
- For Better or For Worse? Learning Minimum Variance Features With Label Augmentation (ICLR 2025 · similarity 0.86)
- Adversarial AutoMixup (ICLR 2024 · similarity 0.86)
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.