Research map / Label-efficient and robust learning / Recognition and long-tail
Model-Aware Contrastive Learning: Towards Escaping the Dilemmas
Read the paperCheck 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 MACL to SimCLR | SimCLR · ImageNet-1K, 200 epochs, batch size 256 | top-1 linear accuracy: 61.9 -> 64.3 | Helped |
| Add MACL to SimCLR | SimCLR · VOC2007 test | AP_all: 49.7 -> 50.1 | Helped |
| Add MACL to SimCSE | SimCSE with RoBERTa · STS12 | Spearman correlation: 70.16 -> 70.76 | Helped |
| Add MACL to SimCSE | SimCSE with RoBERTa · STS13 | Spearman correlation: 81.77 -> 81.43 | Hurt |
| Add MACL to GraphCL | GraphCL · Tox21 | ROC-AUC: 73.87±0.66 -> 74.39±0.29 | Helped |
| Add MACL to GraphCL | GraphCL · BBBP | ROC-AUC: 69.68±0.67 -> 67.98±0.97 | Hurt |
| Add adaptive temperature without reweighting (ablation) | SimCLR loss · ImageNet-1K, 200 epochs, batch size 512, LR-s | top-1 linear accuracy: 64.0 -> 64.9 | Helped |
| Add reweighting without adaptive temperature (ablation) | SimCLR loss · ImageNet-1K, 200 epochs, batch size 512, LR-s | top-1 linear accuracy: 64.0 -> 65.0 | Helped |
| Add adaptive temperature and reweighting (ablation) | SimCLR loss · ImageNet-1K, 200 epochs, batch size 512, LR-s | top-1 linear accuracy: 64.0 -> 65.2 | Helped |
| Add adaptive temperature to reweighted MACL (ablation) | SimCLR loss · ImageNet-100, 200 epochs, batch size 256 | top-1 linear accuracy: 77.32 -> 78.28 | Helped |
| Use MACL rather than NT-Xent | SimCLR · CIFAR10, batch size 64 | linear evaluation accuracy: 82.31 -> 87.11 | Helped |
Closest papers by meaning
- Your contrastive learning problem is secretly a distribution alignment problem (NeurIPS 2024 · similarity 0.89)
- On the Similarities of Embeddings in Contrastive Learning (ICML 2025 · similarity 0.88)
- Aligning Visual Contrastive learning models via Preference Optimization (ICLR 2025 · similarity 0.88)
- Multi-modal contrastive learning adapts to intrinsic dimensions of shared latent variables (NeurIPS 2025 · similarity 0.88)
- A Statistical Theory of Contrastive Learning via Approximate Sufficient Statistics (NeurIPS 2025 · similarity 0.88)
- Self-Supervised Contrastive Learning is Approximately Supervised Contrastive Learning (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.