The abstract is read at the publisher; the summary is JClub's.
Recurrent neural networks trained on a sequential-offer economic choice task reproduce monkey orbitofrontal cortex findings by generating strong order-based choice signals via dynamic value reformatting across delay periods.
- Why it matters: The gap is how neural circuits transform representations when the brain must hold an offer’s value, then compare it with a later offer, despite weak identity-based signals in monkey OFC.
- What they did: The authors trained RNNs to match observed psychometric behavior in an order-based choice task, then used low-dimensional latent-circuit inference to separate feedforward mapping onto a common value scale from asymmetric recurrent interactions.
- The result: Networks reformatted the first offer into a memory signal comparable to the second offer, and phase-to-phase value coding and order-choice preference qualitatively matched monkey OFC data, linking the effect to a dynamical circuit motif.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.