Research map / Label-efficient and robust learning / Semi-supervised learning
SECOS: Semantic Capture for Rigorous Classification in Open-World Semi-Supervised Learning
Hezhao Liu, Jiacheng Yang, Junlong Gao, Mengke Li, Yiqun Zhang, Shreyank N Gowda, Yang Lu
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 Novel Class Semantic Compensation (ablation) | labeled-data-only baseline · CIFAR100 | Novel accuracy: 14.4 -> 80.4 | Improved |
| add Batch-Wise Semantic Recapture alone (ablation) | labeled-data-only baseline · CIFAR100 | Novel accuracy: 14.4 -> 76.7 | Improved |
| add Batch-Wise Semantic Recapture to compensation (ablation) | SECOS · CIFAR100 | All accuracy: 83.9 -> 84.7 | Improved |
| add Novel Class Semantic Compensation (ablation) | labeled-data-only baseline · CUB | All accuracy: 51.2 -> 77.5 | Improved |
| add Batch-Wise Semantic Recapture to compensation (ablation) | SECOS · CUB | All accuracy: 77.5 -> 78.6 | Improved |
| filter pseudo-labels to samples with exactly one candidate label (ablation) | SECOS · CIFAR100 | All accuracy: 82.4 -> 84.7 | Improved |
| filter pseudo-labels to samples with exactly one candidate label (ablation) | SECOS · CUB | All accuracy: 75.3 -> 78.6 | Improved |
| add frozen CLIP teacher instead of EMA self-teaching (ablation) | SECOS · CIFAR100 | All accuracy: 82.4 -> 84.7 | Improved |
| add frozen CLIP teacher instead of EMA self-teaching (ablation) | SECOS · Oxford Flowers | All accuracy: 88.3 -> 90.0 | Improved |
| add frozen CLIP teacher instead of EMA self-teaching (ablation) | SECOS · Oxford Pets | All accuracy: 95.5 -> 95.4 | No clear effect |
| increase adapter downsampling dimension from 2 to 64 (ablation) | SECOS · CIFAR100 and CUB | All accuracy: less than 2% | No clear effect |
| increase batch size from 8 to 64 (ablation) | SECOS · CIFAR100 | All accuracy: 0.35% drop | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SPTNet — SPTNet: An Efficient Alternative Framework for Generalized Category Discovery with Spatial Prompt Tuning (ICLR 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- OwMatch: Conditional Self-Labeling with Consistency for Open-World Semi-Supervised Learning (NeurIPS 2024 · similarity 0.89)
- Learning Textual Prompts for Open-World Semi-Supervised Learning (CVPR 2025 · similarity 0.88)
- OpenLDN: Learning to Discover Novel Classes for Open-World Semi-Supervised Learning (ECCV 2022 · similarity 0.87)
- Targeted Representation Alignment for Open-World Semi-Supervised Learning (CVPR 2024 · similarity 0.85)
- A Unified Framework for Heterogeneous Semi-supervised Learning (CVPR 2025 · similarity 0.84)
- OTMatch: Improving Semi-Supervised Learning with Optimal Transport (ICML 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.
