AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Exploiting Synergistic Cognitive Biases to Bypass Safety in LLMs

Xikang Yang, Biyu Zhou, Xuehai Tang, Jizhong Han, Songlin Hu

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Safety alignment of LLMs · 0 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
add PPO reinforcement learning after SFT (ablation)CognitiveAttack (GPT-4o-mini target) · AdvBenchASR (%): 30.0 -> 64.9Improved
add PPO reinforcement learning after SFT (ablation)CognitiveAttack (Llama-2 target) · AdvBenchASR (%): 36.5 -> 94.3Improved
add PPO reinforcement learning after SFT (ablation)CognitiveAttack (DeepSeek-V3 target) · AdvBenchASR (%): 55.7 -> 94.9Improved
add PPO reinforcement learning after SFT (ablation)CognitiveAttack (Qwen-turbo target) · AdvBenchASR (%): 35.6 -> 88.7Improved
add SFT on bias-annotated rewrites over prompting alone (ablation)CognitiveAttack (DS-v3 target) · AdvBenchASR (%): 41.5 -> 55.7Improved
use learned multi-bias attack instead of PAPCognitiveAttack · HarmBench (30 LLMs)ASR (%): 31.6 -> 60.1Improved
apply paraphrase defenseCognitiveAttack (GPT-4o-mini) · HarmBenchASR (%): 56.5 -> 31.5Hurt
apply retokenization defenseCognitiveAttack (GPT-4o-mini) · HarmBenchASR (%): 56.5 -> 49.3Hurt
apply perplexity filterCognitiveAttack (GPT-4o-mini) · HarmBenchASR (%): 56.5 -> 50.8Hurt
apply LlamaGuard detectorCognitiveAttack (GPT-4o-mini) · HarmBenchASR (%): 56.5 -> 4.5Hurt
apply Perspective API detectorCognitiveAttack (GPT-4o-mini) · HarmBenchASR (%): 56.5 -> 54.8No 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