Artificial intelligencefromInfoQ2 months agoMeta Launches AutoPatchBench to Evaluate LLM Agents on Security FixesAutoPatchBench evaluates LLM agents' ability to autonomously patch security vulnerabilities in C/C++ code.