AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

Persistent Backdoor Attacks Under Continual Fine-Tuning of LLMs

Jing Cui, Yufei Han, Jianbin Jiao, Junge Zhang

AAAI 2026 · Large language models › Instruction and fine-tuning · approach: Safety alignment after fine-tuning

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
optimize trigger tokens for gradient alignment (P-Trojan) vs fixed rare-token trigger (BadNet)backdoor poisoning · SST-2 after cleanup fine-tuningASR (%): 70.00 -> 100.00 (Qwen2.5-0.5B)Improved
gradient-alignment trigger optimization instead of CE-loss trigger optimization (ablation)BadNet-CE · SST-2 after cleanup fine-tuningASR (%): 4.00 -> 100.00 (Qwen2.5-1.5B)Improved
optimize trigger (CE or alignment) instead of non-optimized BadNet (ablation)data-poisoning backdoor at release · SST-2 at releaseASR (%): 46.00 -> 100.00 (Qwen2.5-1.5B)Improved
use BadEdit weight editing instead of data poisoningbackdoor attack · SST-2 at releaseASR (%): 53.00 vs 100.00 (Qwen2.5-1.5B)Hurt
add SST-2 data replay during cross-task fine-tuning (ablation)P-Trojan, Qwen2.5-1.5B · Cross-task fine-tuningSST-2 ASR (%): 67.00 -> 100.00Improved
use FREEZE during cross-task fine-tuning (ablation)P-Trojan, Qwen2.5-1.5B · Cross-task fine-tuningMBPP accuracy (%): 45.80 -> 9.80Hurt
reverse the order of fine-tuning stages (cross-task first, cleanup last) (ablation)P-Trojan, Qwen2.5-1.5B · Fine-tuning orderSST-2 ASR (%): 100.00 -> 99.00No clear effect
change target task from SST-2 to GSM8K (ablation)P-Trojan, Qwen2.5-1.5B · Target task choiceASR / persistence (%): 100.00 -> 100.00No clear effect
continual fine-tune on in-domain SST-5 (ablation)P-Trojan · SST-5 continual fine-tuningASR / persistence (%): 100% ASR, 100% persistenceNo clear effect
add P-Trojan backdoor to the modelQwen2.5-1.5B · Cross-task fine-tuningGSM8K accuracy (%): 58.91 -> 41.00Hurt
add P-Trojan backdoor to the modelQwen2.5-0.5B, LLaMA3.2-1B · Cross-task fine-tuningclean-task accuracy: SST-2 91.59 vs 90.93 (Qwen-0.5B)No 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