Avery Hee-Woon Ryoo

We present POSSM, a novel architecture that combines input cross-attention with a recurrent state-space model, achieving competitive accuracy, fast inference, and efficient generalization for real-time neural decoding applications.