AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Mitigating Safety Fallback in Editing-based Backdoor Injection on LLMs

Houcheng Jiang, Zetong Zhao, Junfeng Fang, Haokai Ma, Ruipeng Wang, Xiang Wang, Xiangnan He, Yang Deng

ICLR 2026 · Adversarial robustness and security › Backdoor and poisoning · approach: Backdoors in large multimodal models

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
add dynamic loss weighting (ablation)DualEdit · DANASR: 71.42 -> 81.51Improved
add dynamic loss weighting (ablation)DualEdit · DANSFR: 36.32 -> 22.67Improved
add dynamic loss weighting (ablation)DualEdit · DNAASR: 66.13 -> 77.93Improved
add dynamic loss weighting (ablation)DualEdit · MisuseSFR: 46.48 -> 40.22Improved
add value anchoring (ablation)DualEdit · DANASR: 75.28 -> 81.51Improved
add value anchoring (ablation)DualEdit · DNASFR: 31.69 -> 27.64Improved
add both dynamic loss weighting and value anchoring (ablation)DualEdit · DNAASR: 56.40 -> 77.93Improved
add both dynamic loss weighting and value anchoring (ablation)DualEdit · DNASFR: 52.13 -> 27.64Improved
add DualEdit backdoor editDualEdit · MMLU, LLaMA-3.1-8Baccuracy: 72.95 -> 71.81No clear effect
add DualEdit backdoor editDualEdit · Average of 7 capability benchmarks, Qwen2.5-7Bavg score: 84.50 -> 81.54Hurt
add DualEdit backdoor editDualEdit · Average of 7 capability benchmarks, LLaMA-2-7Bavg score: 60.56 -> 61.01No clear effect
use DualEdit instead of JailbreakEditDualEdit · DAN, LLaMA-3.1-8BASR with trigger: 75.43% -> 88.07%Improved

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