Research map / Adversarial robustness and security / Attacks on text and graphs
Obscure but Effective: Classical Chinese Jailbreak Prompt Optimization via Bio-Inspired Search
Xun Huang, Simeng Qin, Xiaoshuang Jia, Ranjie Duan, Huanqian Yan, Zhitao Zeng, Fei Yang, Yang Liu, Xiaojun Jia
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 |
|---|---|---|---|
| add multidimensional strategy space to classical Chinese prompts (ablation) | CC-BOS · AdvBench Claude-3.7 | ASR: 18 -> 60 | Improved |
| add bio-inspired optimization on top of strategy space (ablation) | CC-BOS · AdvBench Claude-3.7 | ASR: 60 -> 100 | Improved |
| add translation module to evaluation (ablation) | CC-BOS · AdvBench Claude-3.7 | ASR: 90 -> 100 | Improved |
| CC-BOS vs ICRT under Llama-Guard input and output filters | CC-BOS · AdvBench Claude-3.7 with Llama-Guard | ASR: 26% -> 40.00% | Improved |
| Llama-Guard-3-8B input and output filtering | CC-BOS · AdvBench Claude-3.7 | ASR: 100% -> 40.00% | Hurt |
| CC-BOS vs ICRT on Claude-3.7 | CC-BOS · CLAS 2024 | ASR: 21 -> 99 | Improved |
| CC-BOS vs ICRT on Claude-3.7 | CC-BOS · StrongREJECT-small | ASR: 23.33 -> 98.30 | Improved |
| CC-BOS vs CL-GSO query efficiency | CC-BOS · AdvBench Claude-3.7 | Avg.Q: 21.42 -> 2.38 | Improved |
| transfer prompts across source/target models | CC-BOS · AdvBench transfer | ASR: 76 to 96 across targets | Improved |
Closest papers by meaning
- JailPO: A Novel Black-Box Jailbreak Framework via Preference Optimization Against Aligned LLMs (AAAI 2025 · similarity 0.91)
- Best-of-N Jailbreaking (NeurIPS 2025 · similarity 0.90)
- Systematic Scaling Analysis of Jailbreak Attacks in Large Language Models (NeurIPS 2026 · similarity 0.90)
- JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models (NeurIPS 2024 · similarity 0.89)
- Robust Prompt Optimization for Defending Language Models Against Jailbreaking Attacks (NeurIPS 2024 · similarity 0.89)
- AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models (ICLR 2024 · similarity 0.89)
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.
