AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Purifying Generative LLMs from Backdoors without Prior Knowledge or Clean Reference

Jianwei Li, Jung-Eun Kim

ICLR 2026 · Adversarial robustness and security › Backdoor and poisoning · approach: Backdoors in large multimodal models · 1 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 6 synthetic backdoor variants instead of 1 for signature extraction (ablation)Backdoor signature purification (LoRA) · Targeted Refusal BadNets, LLaMA-2-7B-ChatASR: 40.91 -> 10.66Improved
Add cross-variant alignment term to norm-only scoring (ablation)Backdoor signature scoring · Targeted Refusal BadNets, LLaMA-2-7B-ChatASR / utility: ASR 10.26 -> 10.66; utility 58.86 -> 59.42Mixed
Use norm term instead of alignment-only scoring (ablation)Backdoor signature scoring · Targeted Refusal BadNets, LLaMA-2-7B-ChatASR: 77.04 -> 10.66Improved
Apply signature-guided neuron suppression plus light finetuning vs plain finetuningLLaMA-2-7B-Chat, sentiment steering (full-model) · Sentiment Steering ASR avgASR: 29.96 -> 2.10Improved
Use signature-based suppression instead of CROW at standard learning ratesLLaMA-2-13B-Chat targeted refusal (full-model) · Targeted Refusal ASR avgASR: 68.43 -> 26.46Improved
Include attention heads in signature for Mistral (not only MLP channels) (ablation)Mistral-7B-Instruct LoRA purification · BadNets sentiment steeringASRImproved
Zero out poisoned attention updates while keeping MLP updates (ablation)Poisoned LLaMA-2-7B-Chat · Sanity checkbackdoor persistenceNo clear effect
Zero out poisoned MLP updates while keeping attention updates (ablation)Poisoned LLaMA-2-7B-Chat · Sanity checkbackdoor persistenceImproved
Pruning (magnitude) or quantization as defenseLLaMA-2-13B-Chat · MT-Bench BadNets sentimentMT-Bench: 3.86 -> 3.02 (pruning)Hurt
Plain finetuning on 200 clean samples as defenseLLaMA-2-7B-Chat sentiment (full) · BadNets ASRASR: 59.30 -> 60.0No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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