CheckerBench

CheckerBench was introduced as a benchmark for evaluating whether long-horizon coding agents can synthesize static-analysis checkers, a task requiring agents to interpret a defect specification, inspe…