Research map / Large language models / Language, safety and interpretability
Automatic Dialectic Jailbreak: A Framework for Generating Effective Jailbreak Strategies
Jianghai Yu, Yang Zhou, Zihan Zhou, Lingjuan Lyu, Da Yan, Ruoming Jin, Dejing Dou
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 |
|---|---|---|---|
| use W-ADJ instead of best baseline | jailbreak attack · AdvBench Harmful Behavior | HS: +31.71% | Improved |
| use W-ADJ instead of best baseline | jailbreak attack · AdvBench Harmful Behavior | ASR: +13.9% | Improved |
| use B-ADJ instead of best baseline | jailbreak attack · AdvBench Harmful Behavior | HS: +23.14% | Improved |
| use B-ADJ instead of best baseline | jailbreak attack · AdvBench Harmful Behavior | ASR: +10.29% | Improved |
| apply ADJ under RAIN defense | W-ADJ · AdvBench Harmful Behavior (RAIN) | ASR/HS drop: 0.66% ASR drop, 2% HS drop vs baseline 18.22% and 18.73% | Improved |
| increase number of attack strategies beyond 15 (ablation) | ADJ · AdvBench | ASR and HS: 15 | No clear effect |
| add Hegelian dialectic (defender) to attacker-evaluator self-play (ablation) | ADJ · AdvBench | ASR and HS | Improved |
| change system prompt design (ablation) | ADJ · AdvBench | ASR and HS | No clear effect |
| apply ADJ under perplexity defense | ADJ · AdvBench Harmful Behavior (Perplexity) | ASR and HS | No clear effect |
Closest papers by meaning
- MetaCipher: A Time-Persistent and Universal Multi-Agent Framework for Cipher-Based Jailbreak Attacks for LLMs (AAAI 2026 · similarity 0.89)
- Best-of-N Jailbreaking (NeurIPS 2025 · similarity 0.88)
- Improved Techniques for Optimization-Based Jailbreaking on Large Language Models (ICLR 2025 · similarity 0.88)
- JailPO: A Novel Black-Box Jailbreak Framework via Preference Optimization Against Aligned LLMs (AAAI 2025 · similarity 0.88)
- MT-JailBench: A Modular Benchmark for Understanding Multi-Turn Jailbreak Attacks (NeurIPS 2026 · similarity 0.88)
- Systematic Scaling Analysis of Jailbreak Attacks in Large Language Models (NeurIPS 2026 · similarity 0.88)
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.
