Research map / Label-efficient and robust learning / Semi-supervised learning
Beyond Distribution Estimation: Simplex Anchored Structural Inference Towards Universal Semi-Supervised Learning
Yaxin Hou, Jun Ma, Hanyang Li, Bo Han, Jie Yu, Yuheng Jia
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 auxiliary classification branch for unlabeled data (ablation) | SAGE baseline · CIFAR-10/100 average | average accuracy: 45.56 -> 49.58 | Improved |
| add distribution-agnostic reliability prioritization (DRP) (ablation) | SAGE with AB · CIFAR-10/100 average | average accuracy: 49.58 -> 55.01 | Improved |
| add graph-state relational inference (GRI) (ablation) | SAGE with AB and DRP · CIFAR-10/100 average | average accuracy: 55.01 -> 57.89 | Improved |
| add auxiliary branch (V1 to V2) (ablation) | SAGE baseline · arbitrary unlabeled distribution | pseudo-label accuracy: +13.35 pp | Improved |
| add GRI (V3 to V4) (ablation) | SAGE · arbitrary unlabeled distribution | pseudo-label accuracy and F1: +7.98 pp accuracy, +3.9 pp F1 | Improved |
| use SAGE instead of distribution-estimating CPG under arbitrary unlabeled distribution | UniSSL methods · CIFAR-10 (N=40, M=11650, arbitrary) | accuracy: 50.24 -> 61.24 | Improved |
| fixed simplex ETF anchors instead of no anchor guidance (representation quality vs FreeMatch/CPG) | SAGE · SVHN (gamma_u=150) | inter/intra-class similarity | Improved |
| vary number of anchors K from 50 to 200 (ablation) | SAGE · CIFAR-10 (N=40, M=12390, gamma_u=100) | accuracy: 65.30 (K=50), 63.98 (K=80), 66.83 (K=129), 63.87 (K=200) | No clear effect |
| replace WRN-28-2 backbone with ResNet-50 (2048-d) (ablation) | SAGE · CIFAR-10 (N=40, M=12390, gamma_u=100) | accuracy: SAGE 35.93 vs CPG 24.74 | No clear effect |
| reduce labels to one per class (N=10) | SAGE · CIFAR-10 (M=12390, gamma_u=100) | accuracy: 62.42 -> 43.02 (still above CGMatch 37.31) | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- CGMatch — CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025) · better on 7 of 7 shared results
- FreeMatch — FreeMatch: Self-adaptive Thresholding for Semi-supervised Learning (ICLR 2023) · better on 2 of 2 shared results
- CaliMatch — CaliMatch: Adaptive Calibration for Improving Safe Semi-supervised Learning (ICCV 2025) · better on 1 of 1 shared result
- FixMatch — FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence (NeurIPS 2020) · 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
- Scaling Up Semi-supervised Learning with Unconstrained Unlabelled Data (AAAI 2024 · similarity 0.86)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.86)
- Unlocking the Potential of Unlabeled Data in Semi-Supervised Domain Generalization (CVPR 2025 · similarity 0.85)
- Keep It on a Leash: Controllable Pseudo-label Generation Towards Realistic Long-Tailed Semi-Supervised Learning (NeurIPS 2025 · similarity 0.85)
- InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning (ICML 2024 · similarity 0.85)
- Data Augmentation with Diffusion for Open-Set Semi-Supervised Learning (NeurIPS 2024 · 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.
