AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Broadening the Backdoor Basin: Understanding LLM Backdoors Collapse and Making Backdoors Persistent

Xingyi Zhao, Tian Xie, Xiaojun Qi, Depeng Xu, Shuhan Yuan

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
replace AdamW with BAD-BOOM poisoning (Fisher-ellipsoid SAM)backdoor training · Sentiment Steering VPI, SST-2 SFT, Qwen3-0.6Bpost-SFT ASR: 0.0 -> 97.0Improved
replace AdamW with BAD-BOOM poisoningbackdoor training · Sentiment Steering AddSent, SST-2 SFT, Qwen3-0.6Bpost-SFT ASR: 7.3 -> 97.1Improved
replace AdamW with BAD-BOOM poisoningbackdoor training · Sentiment Steering AddSent, GSM8K SFT, Qwen3-0.6Bpost-SFT ASR: 63.7 -> 93.1Improved
replace vanilla SAM with BAD-BOOM (Fisher-induced ellipsoid) (ablation)backdoor training · Sentiment Steering AddSent, SST-2 SFT, Qwen3-0.6Bpost-SFT ASR: 0.0 -> 97.1Improved
replace vanilla SAM with BAD-BOOM (ablation)backdoor training · Sentiment Steering VPI, SST-2 SFT, Llama3.2-1Bpost-SFT ASR: 0.0 -> 90.5Improved
replace AdamW with vanilla SAMbackdoor training · Sentiment Steering Sleeper, SST-2 SFT, Qwen3-0.6Bpost-SFT ASR: 48.2 -> 95.4Improved
replace AdamW with vanilla SAMbackdoor training · Sentiment Steering AddSent, SST-2 SFT, Qwen3-0.6Bpost-SFT ASR: 7.3 -> 0.0No clear effect
use BAD-BOOM instead of AdamW on a stronger modelbackdoor training · Sentiment Steering AddSent, SST-2 SFT, Qwen3-1.7Bpost-SFT ASR: 96.4 -> 97.1No clear effect
lower poisoning ratio to 1% with BAD-BOOM (ablation)backdoor training · Sentiment Steering Sleeper, SST-2 SFT, Qwen3-0.6Bpost-SFT ASR: 97.1 -> 97.1No clear effect
use BAD-BOOM instead of AdamW at 5% poisoningbackdoor training · Sentiment Steering Sleeper, SST-2 SFT, Qwen3-0.6Bpost-SFT ASR: 0.1 -> 97.1Improved
raise radius rho from 0.001 to 0.005 (ablation)BAD-BOOM · Sentiment Steering VPI, SST-2 SFT, Qwen3-0.6Bpost-SFT ASR: 36.3 -> 97.0Improved
raise radius rho from 0.005 to 0.01 (ablation)BAD-BOOM · Sentiment Steering VPI, SST-2 SFT, Qwen3-0.6Bpost-SFT ASR: 97.0 -> 97.0No clear effect

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