Thats a cool paper from Princeton on LLMs against CTF challenges. Would you surmise that there already exist AI-agents/LLMs that are capable of achieving or surpassing these researchers’ tasks/benchmarks? Such as more capable or more tailored models than off-the-shelf (OTS) GPT4?
(In the case of CTF Challenges) Would the limitation be the capability of the model itself or is this a product of its generalist/OTS nature?
Hey Professor!
Thats a cool paper from Princeton on LLMs against CTF challenges. Would you surmise that there already exist AI-agents/LLMs that are capable of achieving or surpassing these researchers’ tasks/benchmarks? Such as more capable or more tailored models than off-the-shelf (OTS) GPT4?
(In the case of CTF Challenges) Would the limitation be the capability of the model itself or is this a product of its generalist/OTS nature?
I'm not sure, as I haven't been following this space too closely -- but these are the right questions to be asking!