CircuitBuilder: From Polynomials to Circuits via Reinforcement Learning

ArXiv ID: 2603.17075

Summary: CircuitBuilder uses reinforcement learning to discover efficient arithmetic circuits for polynomials. Formulated as a single-player game, an RL agent builds circuits using addition and multiplication gates. The study demonstrates that polynomial circuit synthesis is a compact, verifiable setting for studying self-improving search policies, using SAC and PPO+MCTS.

Read on ArXiv