AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Angel or Demon: Investigating the Plasticity Interventions' Impact on Backdoor Threats in Deep Reinforcement Learning

Oubo Ma, Ruixiao Lin, Yang Dai, Jiahao Chen, Chunyi Zhou, Linkang Du, Shouling Ji

ICML 2026 · Adversarial robustness and security › Backdoor and poisoning · approach: Backdoors in large multimodal models · 0 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
apply SAM during backdoor post-trainingDRL backdoor attacks (TM-Post), robotic tasks · robotic tasksASR: 0.178 -> 0.326Improved
apply SAM during backdoor post-trainingDRL agent benign performance, robotic tasks · robotic tasks TM-PostBTP: 0.745 -> 0.814Improved
apply Weight Clipping in TM-PostDRL backdoor attacks · TM-PostASR: -17.46% averageHurt
apply Spectral Normalization in TM-PostDRL backdoor attacks · TM-PostASR: -11.78% averageHurt
apply interventions during TM-Scratch backdoor trainingDRL backdoor attacks · TM-ScratchASR: Layer Normalization -8.84%No clear effect
apply Layer Normalization in TM-Scratchbenign task performance on sparse-reward tasks · Acrobot and MountainCarBTP: reduces BTP by over 30%Hurt
apply Weight Clipping in TM-Postbenign task performance · TM-PostBTP: -20.19% averageHurt
apply SAM in TM-Post on physics control tasksDRL backdoor attacks · physics control tasksASR: +26.07% relativeImproved
combine SAM with other interventions (Plastic, SLac, SSW)DRL backdoor attacks (TM-Post) · robotic tasksASR: 0.178 -> 0.418 (SSW)Improved
combine interventions (Swiss Cheese, Lac) for mitigationDRL backdoor attacks · TM-PostASRNo clear effect
larger pathological distance among combined interventionsSAM combinations · robotic tasks TM-PostASR: PD 9.43 -> 18.64; ASR 0.368 -> 0.418Improved
inject backdoor during conventional trainingDRL agent training · loss landscape sharpnesssharpness range: range increases by 635.22%Hurt

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