Research map / Adversarial robustness and security / Certified and empirical robustness
Robustness Verification for Contrastive Learning
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 |
|---|---|---|---|
| Use CBC (beta-CROWN for CL) instead of CROWN as the incomplete verifier | RVCL certified instance accuracy · CIFAR-10 CNN-B | certified instance accuracy: 55% vs 34% at eps_test=4/255 | Helped |
| Use CBC instead of CROWN as the incomplete verifier | RVCL verified lower bound tightness · MNIST and CIFAR-10 | gap to PGD upper bound | Helped |
| Use CBC instead of CROWN for ACR computation (ablation) | ACR_CL computation · MNIST and CIFAR-10 | ACR_CL value | No clear effect |
| Increase contrastive adversarial training strength eps_train | CL encoder trained with RoCL · MNIST and CIFAR-10 | ACR_CL and ACR_LE | Helped |
| Increase feature dimension of encoder output (ablation) | CL encoder · CIFAR-10 CNN-B | ACR_CL | No clear effect |
| Increase number of negative samples K (ablation) | ACR_CL estimation · CIFAR-10 CNN-B | ACR_CL | No clear effect |
Closest papers by meaning
- When does Contrastive Learning Preserve Adversarial Robustness from Pretraining to Finetuning? (NeurIPS 2021 · similarity 0.91)
- Robust Pre-Training by Adversarial Contrastive Learning (NeurIPS 2020 · similarity 0.90)
- Investigating Why Contrastive Learning Benefits Robustness against Label Noise (ICML 2022 · similarity 0.89)
- Efficient Adversarial Contrastive Learning via Robustness-Aware Coreset Selection (NeurIPS 2023 · similarity 0.89)
- Weakly Supervised Contrastive Adversarial Training for Learning Robust Features from Semi-supervised Data (CVPR 2025 · similarity 0.87)
- Learning Better Certified Models from Empirically-Robust Teachers (NeurIPS 2026 · 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.