AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Moderate-fitting as a Natural Backdoor Defender for Pre-trained Language Models

Biru Zhu, Yujia Qin, Ganqu Cui, Yangyi Chen, Weilin Zhao, Chong Fu, Yangdong Deng, Zhiyuan Liu, Jingang Wang, Wei Wu, Maosong Sun, Ming Gu

NeurIPS 2022 · Adversarial robustness and security › Backdoor and poisoning · approach: Backdoors in large multimodal models · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (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 changedWhereMeasuredEffect
reduce bottleneck dimension 32 -> 4 (ablation)reparameterized Adapter · SST-2 word-level attackASR (%): 97.15% -> 8.11%Improved
reduce bottleneck dimension 32 -> 4 (ablation)reparameterized Adapter · SST-2 word-level attackACC (%): 94.51% -> 93.68%No clear effect
reduce bottleneck dimension 32 -> 1 (ablation)reparameterized LoRA · SST-2 add-sentence attackASR (%): 100.00 -> 42.21Improved
reduce training epochs 15 -> 1 (ablation)fine-tuning · SST-2 add-sentence attackASR (%): 100.00 -> 35.20Improved
reduce learning rate 2e-5 -> 5e-7 (ablation)fine-tuning · SST-2 style transfer attackASR (%): 86.18 -> 42.21Improved
increase poisoning ratio 10% -> 20% (ablation)reparameterized LoRA dim 1 · SST-2 syntactic attackASR (%): 42.11 -> 65.90Hurt
increase poisoning ratio 2.5% -> 12.5% (ablation)reparameterized LoRA dim 1 · SST-2 word-level attackASR (%): 10.20 -> 14.47Hurt
reduce bottleneck dimension 32 -> 1 on synthetic data (ablation)reparameterized LoRA · Synthetic dataset, SST-2 primary taskACC (%): 94.56 -> 92.20No clear effect
reduce bottleneck dimension 32 -> 1 on synthetic data (ablation)reparameterized LoRA · Synthetic dataset, AG News subsidiary taskACC (%): 95.42 -> 71.21Hurt
reduce learning rate 5e-6 -> 5e-7 on synthetic data (ablation)fine-tuning · Synthetic dataset, AG NewsACC (%): 93.05 -> 69.13Hurt
reduce epochs or learning rate for syntactic attack on AG News (ablation)fine-tuning · AG News syntactic attackASRNo clear effect
compare proposed defense with ONION on syntactic attack (ablation)reduced-capacity LoRA · SST-2 syntactic attackASR (%): 86.29 -> 42.11Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy