Research map / Learning theory and algorithms / Prediction and decision losses
Global Selection of Contrastive Batches via Optimization on Sample Permutations
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 GCBS batch sampling instead of random sampling | UniXcoder · CodeSearchNet / CosQA / AdvTest code search | MRR: +2.2 avg (74.4 to 76.6) | Helped |
| Add GCBS batch sampling instead of random sampling | SimCSE-BERTbase · STS12-STS-B-SICK-R | Spearman: +0.62 | Helped |
| Add GCBS batch sampling instead of random sampling | PromCSE-BERTbase · STS12-STS-B-SICK-R | Spearman: +0.05 | No clear effect |
| Add GCBS batch sampling instead of random sampling | SimCSE-RoBERTabase · STS12-STS-B-SICK-R | Spearman: +0.43 | Helped |
| Add GCBS batch sampling instead of random sampling | PromCSE-RoBERTabase · STS12-STS-B-SICK-R | Spearman: +0.28 | Helped |
| Add GCBS batch sampling instead of random sampling | SimCSE-RoBERTalarge · STS12-STS-B-SICK-R | Spearman: +1.03 | Helped |
| Add GCBS batch sampling instead of random sampling | PromCSE-RoBERTalarge · STS12-STS-B-SICK-R | Spearman: +0.27 | Helped |
| Add GCBS batch sampling instead of random sampling | SimCLR · Imagenette kNN | test accuracy: +1.7 | Helped |
| Add GCBS batch sampling instead of random sampling | MoCo · Imagenette kNN | test accuracy: +1.3 | Helped |
| Add GCBS batch sampling instead of random sampling | SimCLR · Cifar10 kNN | test accuracy: +2.3 | Helped |
| Add GCBS batch sampling instead of random sampling | MoCo · Cifar10 kNN | test accuracy: +0.3 | Helped |
| Add GCBS batch sampling instead of random sampling (ablation) | UniXcoder training on CodeSearchNet Ruby · Global minus training loss gap | relative gap reduction: gap reduced by 40% | Helped |
Closest papers by meaning
- On the Similarities of Embeddings in Contrastive Learning (ICML 2025 · similarity 0.89)
- Breaking the Batch Barrier (B3) of Contrastive Learning via Smart Batch Mining (NeurIPS 2025 · similarity 0.89)
- EMC$^2$: Efficient MCMC Negative Sampling for Contrastive Learning with Global Convergence (ICML 2024 · similarity 0.86)
- Uncertainty-Aware Contrastive Learning with Hard Negative Sampling for Code Search Tasks (AAAI 2025 · similarity 0.86)
- Does GCL Need a Large Number of Negative Samples? Enhancing Graph Contrastive Learning with Effective and Efficient Negative Sampling (AAAI 2025 · similarity 0.86)
- A Statistical Theory of Contrastive Learning via Approximate Sufficient Statistics (NeurIPS 2025 · similarity 0.86)
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.