10.1101/2025.03.13.642273

Neurocomputational underpinnings of suboptimal beliefs in recurrent neural network-based agents