AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Uncovering Hidden Triggers: Backdoor Attribution in Language Models

Miao Yu, Zhenhong Zhou, Moayad Aloqaily, Kun Wang, Biwei Huang, Stephen Wang, Yueming Jin, Qingsong Wen

ICML 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
ablate the top-16 ACIE backdoor attention heads (ablation)backdoored Llama-2-7B-chat · agnews sentenceASR (%): 100.0 -> 29.03Improved
ablate the top-16 heads on jailbreak backdoor (ablation)backdoored Llama-2-7B-chat · harmful randomASR (%): 75.78 -> 7.81Improved
ablate only 1-8 heads (ablation)backdoored Llama-2-7B-chat · alpaca beginASR (%): 100.0 -> 99.22 (n=8)No clear effect
ablate top-32 heads on fixed-output backdoor (ablation)backdoored Qwen-2.5-7B-Instruct · alpaca beginASR (%): 100.0 -> 90.62No clear effect
ablate top-32 heads on label modification backdoor (ablation)backdoored Qwen-2.5-7B-Instruct · agnews sentenceASR (%): 91.94 -> 30.65Improved
add backdoor vector to hidden state of clean inputbackdoor vector · Llama-2-7B-chat jailbreakASR (%): 0.00 -> 85.16Improved
subtract backdoor vector from hidden state of triggered inputbackdoor vector · Llama-2-7B-chat jailbreakASR (%): 75.78 -> 0.39Improved
subtract backdoor vector on Qwen jailbreak backdoorbackdoor vector · Qwen-2.5-7B-Instruct jailbreakASR (%): 78.91 -> 55.86Mixed
build the vector from random heads instead of top-ACIE heads (ablation)backdoor vector · Llama-2-7B-chatASR (%): 100.0 -> 92.29 (random) vs 1.61 (attributed)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