Research map / Label-efficient and robust learning / Self-supervised and contrastive
Rethinking Minimal Sufficient Representation in Contrastive Learning
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 input reconstruction regularization | SimCLR · DTD transfer classification | accuracy (%): 29.52 -> 33.67 | Helped |
| Add an InfoNCE lower-bound estimate of input-representation mutual information | SimCLR · DTD transfer classification | accuracy (%): 29.52 -> 34.52 | Helped |
| Add input reconstruction regularization | SimCLR · CIFAR10 classification | accuracy (%): 85.76 -> 85.78 | No clear effect |
| Add an InfoNCE lower-bound estimate of input-representation mutual information | SimCLR · CIFAR10 classification | accuracy (%): 85.76 -> 85.45 | Hurt |
| Add input reconstruction regularization | SimCLR · CUBirds transfer classification | accuracy (%): 8.87 -> 10.89 | Helped |
| Add input-representation mutual-information regularization | SimCLR · VOC07+12 object detection | AP: 45.5 -> 48.1; +2.6 | Helped |
| Add input-representation mutual-information regularization | SimCLR · COCO object detection | APbb: 38.6 -> 39.3; +0.7 | Helped |
| Add input-representation mutual-information regularization | SimCLR · COCO instance segmentation | APmk: 33.9 -> 34.5; +0.6 | Helped |
| Add inverse predictive loss to align representations from different views (ablation) | SimCLR · CIFAR10 classification | accuracy (%): 85.76 -> 85.86 | No clear effect |
| Add inverse predictive loss to align representations from different views (ablation) | SimCLR · DTD transfer classification | accuracy (%): 29.52 -> 30.15 | Helped |
| Add multi-view information bottleneck regularization to align view representations (ablation) | SimCLR with Gaussian noise · DTD transfer classification | accuracy (%): 31.70 -> 31.17 | Hurt |
| Increase the amount of input-representation mutual-information regularization (ablation) | SimCLR · Downstream classification | accuracy | Mixed |
Closest papers by meaning
- What Makes for Good Views for Contrastive Learning? (NeurIPS 2020 · similarity 0.90)
- On the Similarities of Embeddings in Contrastive Learning (ICML 2025 · similarity 0.88)
- A Statistical Theory of Contrastive Learning via Approximate Sufficient Statistics (NeurIPS 2025 · similarity 0.88)
- Separating common from salient patterns with Contrastive Representation Learning (ICLR 2024 · similarity 0.88)
- Self-Supervised Contrastive Learning is Approximately Supervised Contrastive Learning (NeurIPS 2025 · similarity 0.88)
- Multi-modal contrastive learning adapts to intrinsic dimensions of shared latent variables (NeurIPS 2025 · 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.