Research map / Adversarial robustness and security / Attacks on text and graphs
SlotGCG: Exploiting the Positional Vulnerability in LLMs for Jailbreak Attacks
Seungwon Jeong, Jiwoo Jeong, Hyeonjin Kim, Yunseok Lee, Woojin Lee
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 SlotGCG slot selection | GCG · AdvBench (6 models avg) | ASR (%): 66.7% -> 80.0% | Improved |
| add SlotGCG slot selection | AttnGCG · AdvBench (6 models avg) | ASR (%): 61.7% -> 86.3% | Improved |
| add SlotGCG slot selection | I-GCG · AdvBench (6 models avg) | ASR (%): 73.0% -> 85.7% | Improved |
| add SlotGCG slot selection | GCG-Hij · AdvBench (6 models avg) | ASR (%): 78.0% -> 84.3% | Improved |
| add SlotGCG slot selection | GBDA · AdvBench (6 models avg) | ASR (%): 20.3% -> 40.0% | Improved |
| add SlotGCG slot selection | GCG on Mistral-7B · AdvBench | ASR (%): 86.0% -> 86.0% | No clear effect |
| add SlotGCG slot selection | AttnGCG on Vicuna-7B · AdvBench | ASR (%): 88.0% -> 82.0% | Hurt |
| add SlotGCG slot selection | AttnGCG on Llama-2-13B · AdvBench | ASR (%): 20.0% -> 82.0% | Improved |
| add SlotGCG slot selection under defenses | GCG · AdvBench with 9 defenses (avg) | ASR (%): 18.9% -> 46.2% | Improved |
| add SlotGCG slot selection under defenses | I-GCG · AdvBench with 9 defenses (avg) | ASR (%): 21.8% -> 53.6% | Improved |
| add SlotGCG slot selection under Perplexity Filter | GCG · AdvBench Perplexity Filter | ASR (%): 0.0% -> 0.0% | No clear effect |
| add SlotGCG slot selection | GCG · AdvBench (6 models avg) | iterations to success: 72.59 -> 28.59 | Improved |
Closest papers by meaning
- 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 0.89)
- Bag of Tricks: Benchmarking of Jailbreak Attacks on LLMs (NeurIPS 2024 · similarity 0.88)
- Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks (ICLR 2025 · similarity 0.88)
- JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models (NeurIPS 2024 · similarity 0.88)
- One Model Transfer to All: On Robust Jailbreak Prompts Generation against LLMs (ICLR 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.
