Reproduce “CT0: Cybersecurity Transformer” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Víctor Mayoral-Vilches
CT0 combines a typed decision head with a generative cybersecurity agent, routing uncertain cases to text generation. The head selects among options during prompt processing without decoding answer tokens. On Cybench-Subtask, it achieves 21.1% (36/171) exact-match answers across 171 subtasks from 40 CTF challenges, versus 7.6% for its Qwen3.5-4B base under a fixed one-attempt budget. This supports improvement on this benchmark, not professional-level performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.