QiMeng-SALV introduces a novel framework for Verilog code generation that shifts reinforcement learning optimization from module-level to signal-level rewards. By leveraging AST analysis and ...