A Dual-Ensemble Model of Infralimbic Cortex Function in Behavioural Flexibility

The infralimbic cortex (IL) has been described as both facilitator and terminator of reward-seeking, yet both functions appear mutually exclusive. We reconcile this discrepancy by proposing a dual-ensemble model: one for action-outcome prediction and another for prediction correction. Using longitudinal miniscope imaging in rats, we show that during training, the action-outcome predictor ensemble sharpens its activity by suppressing the majority of the IL neuronal population. As contingencies stabilize, entropy decreases, reflecting a refined predictive model. During extinction, the prediction correction ensemble activates in response to reward omission. A computational simulation based on t