Abstract: Neural theorem provers rely on computationally expensive search algorithms to recover from failed proof attempts, lacking the human ability to anticipate and avoid unproductive paths. We ...