Research map / Label-efficient and robust learning / Recognition and long-tail
Joint Contrastive Learning with Infinite Possibilities
Qi Cai, Yu Wang, Yingwei Pan, Ting Yao, Tao Mei
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 |
|---|---|---|---|
| Use JCL instead of MoCo v2 pre-training | ResNet-50 linear classifier · ImageNet1K | top-1 accuracy: 67.5 -> 68.7 | Improved |
| Use JCL instead of MoCo v2 pre-training | ResNet-50 linear classifier · ImageNet1K | top-5 accuracy: 88.0 -> 89.0 | Improved |
| Use JCL instead of MoCo v2 pre-training | Faster R-CNN with ResNet-50 · MS COCO object detection | AP bbox: 37.6 -> 38.1 | Improved |
| Use JCL instead of MoCo v2 pre-training | Mask R-CNN with ResNet-50 · MS COCO instance segmentation | AP mask: 35.3 -> 35.6 | Improved |
| Use JCL instead of MoCo v2 pre-training | Keypoint R-CNN with ResNet-50 · MS COCO keypoint detection | AP keypoint: 66.0 -> 66.2 | Improved |
| Use JCL instead of MoCo v2 pre-training | Keypoint R-CNN with ResNet-50 · MS COCO keypoint detection | AP50 keypoint: 87.2 -> 87.2 | No clear effect |
| Vary the number of positive keys M' (ablation) | JCL · ImageNet100 linear classification | classification accuracy | No clear effect |
| Vary covariance strength lambda (ablation) | JCL · ImageNet100 linear classification | classification accuracy | No clear effect |
| Increase covariance strength lambda beyond the reported stable range (ablation) | JCL · ImageNet100 linear classification | classification accuracy | Hurt |
| Vary softmax temperature tau (ablation) | JCL · ImageNet100 linear classification | classification accuracy | Mixed |
Closest papers by meaning
- On the Similarities of Embeddings in Contrastive Learning (ICML 2025 · similarity 0.89)
- Enhancing Contrastive Learning Inspired by the Philosophy of “The Blind Men and the Elephant” (AAAI 2025 · similarity 0.89)
- Self-Supervised Contrastive Learning is Approximately Supervised Contrastive Learning (NeurIPS 2025 · similarity 0.88)
- Generalization Analysis for Supervised Contrastive Representation Learning under Non-IID Settings (ICML 2025 · similarity 0.87)
- Variational Supervised Contrastive Learning (NeurIPS 2025 · similarity 0.87)
- Bridging Mini-Batch and Asymptotic Analysis in Contrastive Learning: From InfoNCE to Kernel-Based Losses (ICML 2024 · similarity 0.87)
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.
