Binary Exploitation Benchmark

An internal benchmark developed by Anthropic's frontier red team to evaluate the autonomous offensive capabilities of large language models in binary exploitation, covering advanced cybersecurity tasks such as control-flow hijacking.