Brain-inspired synaptic transistors for in-situ spiking reinforcement learning with eligibility trace

Brain-inspired reinforcement learning (RL) represents a pivotal pathway toward artificial general intelligence, yet existing hardware implementations based on artificial neural networks lack critical biological mechanisms like third-terminal modulated eligibility traces and dynamic reward signaling. Emerging materials can address these challenges by mimicking RL’s complex dynamics with revolutionary efficiency. Here we demonstrate a brain-inspired SNN-based RL computing architecture using α-In ₂ Se ₃ ferroelectric semiconductor field-effect transistor (FeS-FET). By leveraging the intrinsic in-plane and out-of-plane polarization coupling of α-In ₂ Se ₃ , the multi-terminal conductance modulation in the FeS-FET enables reward signal modulation of RL. The ferroelectric relaxation is utilized to implement biological eligibility trace decay, thereby enhancing the algorithm's processing capability. autonomous driving tasks are then demonstrated with RL neural network constructed by the α-In ₂ Se ₃ FeS-FET array, where in-situ reward-based weight updates and eligibility trace decay are performed without any external memory or computing units. Our solution paves the way for a SNN-based RL computing architecture with full functionality, low energy consumption and reduced hardware overhead.