AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Cost-Minimized Label-Flipping Poisoning Attack to LLM Alignment

Shigeki Kusaka, Keita Saito, Mikoto Kudo, Takumi Tanabe, Akifumi Wachi, Youhei Akimoto

AAAI 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: LLM jailbreak and injection defense · 3 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
post-process the attack with PCM (convex cost minimization)RLHFPoison · HH-RLHF, Phi-3.5-miniflip rate: -30.4%Improved
post-process the attack with PCMRLHFPoison · HH-RLHF, Llama-2-13bflip rate: -20.0%Improved
post-process the attack with PCMRLHFPoison · PKU-SafeRLHF, Llama-2-7bflip rate: -10.6%Improved
post-process the attack with PCMRLHFPoison · HH-RLHF, Llama-2-13boutput length increase: 1.63 -> 1.27Hurt
post-process the attack with PCMRLHFPoison · PKU-SafeRLHF, Llama-2-13boutput length increase: 0.25 -> 0.37Improved
post-process the attack with PCMRLHFPoison · PKU-SafeRLHF, Llama-2-7boutput length increase: 0.29 -> 0.29No clear effect
apply PCM when feature dimension exceeds dataset sizeRLHFPoison · Social-Reasoning-RLHF (N=3,820)flip rateNo clear effect
increase dataset size NPCM · synthetic datacost reductionImproved
increase annotation granularity m (ablation)PCM discretization · synthetic dataperformance loss rateImproved

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