Research map / Label-efficient and robust learning / Semi-supervised learning
Language Semantic Graph Guided Data-Efficient Learning
Wenxuan Ma, Shuang Li, Lincan Cai, Jingxuan Kang
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 language semantic graph guided training (LSG) to vanilla fine-tuning | ResNet-50 fine-tuning · FGVC Aircraft 15% labels | accuracy (%): 41.6 -> 55.6 | Improved |
| add regularization loss L_r (ablation) | LSG · FGVC Aircraft 15% labels | accuracy (%): 44.2 -> 55.6 | Improved |
| add alignment loss L_align (ablation) | LSG · FGVC Aircraft 15% labels | accuracy (%): 50.7 -> 55.6 | Improved |
| use trainable classifier instead of fixed language head (ablation) | LSG · FGVC Aircraft 15% labels | accuracy (%): 47.9 -> 55.6 | Improved |
| add stop gradient in L_r (ablation) | LSG · FGVC Aircraft 15% labels | accuracy (%): 48.3 -> 55.6 | Improved |
| replace GCN-based alignment with text prototype alignment (ablation) | LSG · FGVC Aircraft 15% labels | accuracy (%): 55.6 -> 51.1 | Hurt |
| replace GCN-based alignment with contrastive alignment (ablation) | LSG · FGVC Aircraft 15% labels | accuracy (%): 55.6 -> 54.2 | Hurt |
| refine label embeddings with GCN (ablation) | label embeddings · FGVC Aircraft | Calinski-Harabasz index: 86.5 -> 1364.1 | Improved |
| substitute GAT for GCN (ablation) | LSG · FGVC Aircraft SSL | accuracy | Mixed |
| vary edge-connection threshold tau (cross-label edge ratio) (ablation) | LSG · image datasets | primary model accuracy | Mixed |
| extend LSG with unlabeled data (semi-supervised) | LSG · FGVC Aircraft 15% labels | accuracy (%): 55.6 -> 71.4 | Improved |
| add LSG to Tip-Adapter style CLIP tuning | CLIP ViT-B · FGVC Aircraft 15% labels | accuracy (%): 36.9 -> 48.9 | Improved |
Closest papers by meaning
- Harnessing Language Model for Cross-Heterogeneity Graph Knowledge Transfer (AAAI 2025 · similarity 0.86)
- Efficient and Context-Aware Label Propagation for Zero-/Few-Shot Training-Free Adaptation of Vision-Language Model (ICLR 2025 · similarity 0.86)
- Scale-Free Graph-Language Models (ICLR 2025 · similarity 0.85)
- Leveraging Large Language Models for Node Generation in Few-Shot Learning on Text-Attributed Graphs (AAAI 2025 · similarity 0.85)
- SSTAG: Structure-Aware Self-Supervised Learning Method for Text-Attributed Graphs (NeurIPS 2025 · similarity 0.85)
- Toward a Unified Statistical Theory of Unsupervised Pretraining and Supervised Neural Knowledge Graph Learning (NeurIPS 2026 · similarity 0.85)
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.
