AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Backdooring Rationalization

Lingxiao Kong, Jiahui Jiang, Wenchao Xu, Lei Wu

AAAI 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
use tailored BadRNL attack instead of naive BadNet/CbatGRU rationalization model · Fever token-levelASR (%): 47.07 -> 100.00 (vs BadNet)Improved
use tailored BadRNL attack instead of naive BadNetGRU rationalization model · MultiRC sentence-levelASR (%): 87.93 -> 100.00Improved
apply AttDef defense to BadRNLGRU rationalization model · MultiRC sentence-levelASR (%): 100.00 -> 94.37No clear effect
apply adversarial training defense to BadRNLGRU rationalization model · Fever sentence-levelASR (%): 100.00 -> 89.10Mixed
add BadRNL backdoor to clean modelGRU rationalization model · Fever sentence-levelclean accuracy (%): 69.65 -> 70.31No clear effect
add BadRNL backdoor to clean modelGRU rationalization model · Fever sentence-levelgold rationale F1 (%): 42.33 -> 43.06No clear effect
add rationale loss L_RAT (ablation)BadRNL · Fever/MultiRCgold rationale F1: up to 3.28%Improved
use gradient-searched triggers instead of random triggers (ablation)BadRNL · FeverASRImproved
insert trigger at rationale position instead of beginning or ending (ablation)BadRNL · FeverASRImproved
use poison-priority sampling instead of uniform sampling (ablation)BadRNL · five datasetsASRImproved
increase poisoning rate beyond a threshold (ablation)BadRNL · Fever/MultiRCACC and GRHurt
raise poisoning rate to 0.05 (ablation)BadRNL · FeverASRImproved

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