LessWrong AI
· Communities
Ten Thousand Cyber Labs for Training & Eval
Multiple recent developments - such as GPT-5.6 hacking into HuggingFace to cheat in a cybersecurity eval - have underscored the need to increase our capability to evaluate the cybersecurity capabilities of new and upcoming AI models.TarantuBench-v2 aims to do two things:Evaluate the cybersecurity capabilities of new an