Research map / Label-efficient and robust learning / Recognition and long-tail
Quality-Aware Self-Training on Differentiable Synthesis of Rare Relational Data
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 |
|---|---|---|---|
| Add Phase 2 fine-tuning with the stronger BLSTM classifier (ablation) | QAST · UCI car, ecoli, glass, pageblocks, satimage, thyroid | Macro-F1: 91.98 -> 94.96; 77.35 -> 84.70; 62.12 -> 71.55; 74.47 -> 85.96; 87.72 -> 89.60; 90.59 -> 90.88 | Mixed |
| Use a self-adaptive discriminator confidence threshold instead of fixed Ps (ablation) | QAST w/o ft · UCI car, ecoli, glass | Macro-F1: 91.45; 77.70; 64.10 | Helped |
| Use a self-adaptive classifier committee vote threshold instead of fixed Tl (ablation) | QAST w/o ft · UCI car, ecoli, glass | Macro-F1: 90.90; 87.49; 88.85; 78.13; 79.20; 77.87; 60.86; 63.38; 61.62 | Helped |
| Use random, SMOTE, and Mixup together to generate generator seeds (ablation) | QAST w/o ft · UCI car, ecoli, glass | Macro-F1: 91.52; 91.29; 90.59; 88.16; 78.34; 76.45; 74.46; 79.02; 68.99; 59.18; 62.71; 63.65 | Helped |
| Vary the classifier committee composition (ablation) | QAST w/o ft · UCI car, ecoli, glass | Macro-F1: 93.39; 94.10; 94.26; 95.06; 79.43; 80.21; 78.66; 80.05; 72.06; 68.63; 72.08; 72.30 | No clear effect |
| Use QAST w/o ft rather than the strongest reported baseline | QAST w/o ft · NASA JM1 | Macro-F1: 59.88 -> 57.02 | Hurt |
Closest papers by meaning
- A Unified Generative Adversarial Network Training via Self-Labeling and Self-Attention (ICML 2021 · similarity 0.89)
- Self-Supervised GANs with Label Augmentation (NeurIPS 2021 · similarity 0.88)
- Augmentation-Aware Self-Supervision for Data-Efficient GAN Training (NeurIPS 2023 · similarity 0.88)
- Fusing Conditional Submodular GAN and Programmatic Weak Supervision (AAAI 2024 · similarity 0.87)
- Taming the Tail in Class-Conditional GANs: Knowledge Sharing via Unconditional Training at Lower Resolutions (CVPR 2024 · similarity 0.86)
- CG-TGAN: Conditional Generative Adversarial Networks with Graph Neural Networks for Tabular Data Synthesizing (AAAI 2025 · 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.