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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| post-process the attack with PCM (convex cost minimization) | RLHFPoison · HH-RLHF, Phi-3.5-mini | flip rate: -30.4% | Improved |
| post-process the attack with PCM | RLHFPoison · HH-RLHF, Llama-2-13b | flip rate: -20.0% | Improved |
| post-process the attack with PCM | RLHFPoison · PKU-SafeRLHF, Llama-2-7b | flip rate: -10.6% | Improved |
| post-process the attack with PCM | RLHFPoison · HH-RLHF, Llama-2-13b | output length increase: 1.63 -> 1.27 | Hurt |
| post-process the attack with PCM | RLHFPoison · PKU-SafeRLHF, Llama-2-13b | output length increase: 0.25 -> 0.37 | Improved |
| post-process the attack with PCM | RLHFPoison · PKU-SafeRLHF, Llama-2-7b | output length increase: 0.29 -> 0.29 | No clear effect |
| apply PCM when feature dimension exceeds dataset size | RLHFPoison · Social-Reasoning-RLHF (N=3,820) | flip rate | No clear effect |
| increase dataset size N | PCM · synthetic data | cost reduction | Improved |
| increase annotation granularity m (ablation) | PCM discretization · synthetic data | performance loss rate | Improved |
Closest papers by meaning
- Safety Alignment Should be Made More Than Just a Few Tokens Deep (ICLR 2025 · similarity 0.89)
- Improving LLM Safety Alignment with Dual-Objective Optimization (ICML 2025 · similarity 0.88)
- Safety Layers in Aligned Large Language Models: The Key to LLM Security (ICLR 2025 · similarity 0.88)
- Fail-Closed Alignment for Large Language Models (NeurIPS 2026 · similarity 0.87)
- Mission Impossible: A Statistical Perspective on Jailbreaking LLMs (NeurIPS 2024 · similarity 0.87)
- Is Poisoning a Real Threat to DPO? Maybe More So Than You Think (AAAI 2025 · similarity 0.87)
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.
