llm绕过对抗
Quote:
Originally Posted by
santrax
- GPT6 Sol (be careful with the prompt)
- Opus 4.6 and Opus 4.7. Patch the .exe to disable the prompt that prevents security attacks.
- Kimi K3
- Any abliterated local model.
What kind of prompting do you use for Sol to prevent it from refusing something or getting flagged?
I sometimes ask like I am trying to work on malware analysis. Bascially, I replace teh Anti cheat with Malware everywhre in my prompt. That has appeared to work so far for fixing some kvm/qemu VM cheks..
Im on OpenAI Pro 100 (5x) plan.