Research map / Adversarial robustness and security / Attacks on text and graphs
Adversarially trained neural representations are already as robust as biological neural representations
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 |
|---|---|---|---|
| Adversarial training (l2 eps=3) of ResNet50 and WideResNet50-4 | ImageNet unit-level penultimate layer · ImageNet unit-level adversarial sensitivity | normalized adversarial sensitivity (lower is better): slightly less sensitive than IT neural sites | Helped |
| Standard (non-robust) training of ResNet50 | ImageNet unit-level penultimate layer · ImageNet unit-level adversarial sensitivity | normalized adversarial sensitivity (lower is better): over 10-fold larger than baseline variation, ~10-fold more than IT sites | Hurt |
| Adversarial perturbations on non-preferred images with l2 eps=10 | IT neural sites · category preference manipulation | multi-unit firing rate (Hz): 310±60Hz vs 216±60Hz for most-preferred categories | Helped |
Closest papers by meaning
- Biologically Inspired Mechanisms for Adversarial Robustness (NeurIPS 2020 · similarity 0.92)
- Simulating a Primary Visual Cortex at the Front of CNNs Improves Robustness to Image Perturbations (NeurIPS 2020 · similarity 0.88)
- Modeling Biological Immunity to Adversarial Examples (CVPR 2020 · similarity 0.88)
- Towards robust vision by multi-task learning on monkey visual cortex (NeurIPS 2021 · similarity 0.88)
- Enhancing Adversarial Robustness in SNNs with Sparse Gradients (ICML 2024 · similarity 0.85)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.84)
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.