Research map / Label-efficient and robust learning / Recognition and long-tail
SNAPHARD CONTRAST LEARNING
Changpu Meng, Jie Yang, Wanqing Li, Yi Guo
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 |
|---|---|---|---|
| replace SupCon/VarCon loss with SPACL hard positive/negative contrast | ResNet-50 supervised · CIFAR-100 | top-1 accuracy: 78.29 -> 80.84 (vs VarCon) | Improved |
| replace VarCon with SPACL | ResNet-50 supervised · ImageNet-1K | top-1 accuracy: 79.36 -> 80.98 | Improved |
| use SPACL in self-supervised training vs best baseline | ResNet-50 · ImageNet-1K | top-1 accuracy: 74.28 -> 76.59 (vs BYOL) | Improved |
| use SPACL in weakly-supervised training | ResNet-50 · ImageNet-1K | top-1 accuracy: 19.08 -> 22.08 (vs MaskCon) | Improved |
| use SPACL for multimodal KG link prediction | SPACL · FB15K-237 | MRR: 0.400 -> 0.413 (vs LMKGE) | Improved |
| use SPACL for OOD intent detection | BERT-base · BANKING 75% IND | F1-OOD: 74.94 -> 75.72 (vs TBOS) | No clear effect |
| swap contrastive loss in MoCo v3 setup for SPACL | ViT-S/16 · ImageNet-1K | top-1 accuracy: 72.5 -> 74.3 | Improved |
| add negative screening (relative threshold for hard negatives) (ablation) | SPACL · CIFAR-100 and FB15K-237 | accuracy / link metrics | Improved |
| add adversarial negatives (ablation) | SPACL · CIFAR-100 and FB15K-237 | accuracy / link metrics | Improved |
| add anchor selection (ablation) | SPACL · CIFAR-100 and FB15K-237 | accuracy / link metrics | Improved |
| add hard positive selection (farthest-point) (ablation) | SPACL · CIFAR-100 and FB15K-237 | accuracy / link metrics | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MaskCon — MaskCon: Masked Contrastive Learning for Coarse-Labelled Dataset (CVPR 2023) · better on 2 of 2 shared results
- Barlow Twins — Barlow Twins: Self-Supervised Learning via Redundancy Reduction (ICML 2021) · 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
- Decoupled Contrastive Learning (ECCV 2022 · similarity 0.88)
- On the Similarities of Embeddings in Contrastive Learning (ICML 2025 · similarity 0.87)
- A Statistical Theory of Contrastive Learning via Approximate Sufficient Statistics (NeurIPS 2025 · similarity 0.86)
- EMC$^2$: Efficient MCMC Negative Sampling for Contrastive Learning with Global Convergence (ICML 2024 · similarity 0.85)
- Continual Multimodal Contrastive Learning (NeurIPS 2025 · similarity 0.85)
- $\mathbb{X}$-Sample Contrastive Loss: Improving Contrastive Learning with Sample Similarity Graphs (ICLR 2025 · 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.
