Combining Trained Models in Reinforcement Learning

ArXi:2605.02159v1 Announce Type: new Deep reinforcement learning (DRL) has delivered strong results in domains such as Atari and Go, but it still suffers from high sample cost and weak transfer beyond the