Research map / Large language models / Language, safety and interpretability
EquaCode: A Multi-Strategy Jailbreak Approach for Large Language Models via Equation Solving and Code Completion
Zhen Liang, Hai Huang, Zhengkui Chen
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 equation module on top of natural-language step decomposition (STSA) (ablation) | jailbreak prompts · AdvBench subset 50, LLaMA-ASR, 6-LLM average | ASR (%): 17.33 -> 44.67 | Improved |
| add code-completion wrapper to STSA (ablation) | jailbreak prompts · AdvBench subset 50, LLaMA-ASR, 6-LLM average | ASR (%): 17.33 -> 65.73 | Improved |
| add code module to equation module (ablation) | EquaCode · AdvBench subset 50, LLaMA-ASR, 6-LLM average | ASR (%): 44.67 -> 87.33 | Improved |
| add equation module to code module (ablation) | EquaCode · AdvBench subset 50, LLaMA-ASR, 6-LLM average | ASR (%): 65.73 -> 87.33 | Improved |
| add equation module to STSA on GPT-3.5-turbo (ablation) | GPT-3.5-turbo · AdvBench subset 50, LLaMA-ASR | ASR (%): 76.00 -> 74.00 | Hurt |
| add equation module to STSA on GPT-4o and GPT-4o-mini (ablation) | GPT-4o, GPT-4o-mini · AdvBench subset 50, LLaMA-ASR | ASR (%): 0 -> 30 (GPT-4o), 0 -> 16 (GPT-4o-mini) | Improved |
| EquaCode versus FlipAttack on Llama-3.1-405B | Llama-3.1-405B · AdvBench (520, GPT-ASR) | ASR (%): 27.50 -> 17.88 | Hurt |
| EquaCode versus CodeChameleon on GPT-4o | GPT-4o · AdvBench (520, GPT-ASR) | ASR (%): 92.67 -> 87.12 | Hurt |
| use EquaCode against Llama Guard 7B moderation | Llama Guard 7B · defense | bypass rate (%): 100% | Improved |
| apply output filtering defense against EquaCode | EquaCode · defense | attack success rate (%): 14% success under output filter | Hurt |
Later papers that compared with this method
1 later accepted paper reports EquaCode as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- CodeMimicry: Exploiting Safety Generalization Lag in Large Language Models via Structured Code Completion (NeurIPS 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FlipAttack — FlipAttack: Jailbreak LLMs via Flipping (ICML 2025) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- An Interpretable N-gram Perplexity Threat Model for Large Language Model Jailbreaks (ICML 2025 · similarity 0.92)
- JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models (NeurIPS 2024 · similarity 0.91)
- Systematic Scaling Analysis of Jailbreak Attacks in Large Language Models (NeurIPS 2026 · similarity 0.91)
- Improved Techniques for Optimization-Based Jailbreaking on Large Language Models (ICLR 2025 · similarity 0.90)
- JailPO: A Novel Black-Box Jailbreak Framework via Preference Optimization Against Aligned LLMs (AAAI 2025 · similarity 0.90)
- Multi-Turn Jailbreaking Large Language Models via Attention Shifting (AAAI 2025 · similarity 0.90)
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.
