Beliefs about states of the world profoundly impact decision-making and learning, but little is known about how neural circuits represent and update beliefs. We performed projection-specific recordings and perturbations from neurons in the orbitofrontal cortex (OFC) projecting to the intermediate or rostral caudate putamen (CPi/CPr) in rats performing a task with hidden reward states. Stimulating OFC-CPi neurons biased rats’ beliefs towards high reward states. Recordings from optogenetically-tagged OFC-CPi neurons showed that they encoded evidence for high reward states, and evidence encoding was shaped by local inhibition. Finally, projection-specific perturbations disrupted encoding of hidden states within OFC via long-range cortico-basal ganglia-thalamic loops. These findings reveal the circuit implementation of a core cognitive computation, updating subjective beliefs about abstract latent states of the environment.
Measuring and Exploiting Confirmation Bias in LLM-Assisted Security Code Review
arXiv:2603.18740v1 Announce Type: cross Abstract: Security code reviews increasingly rely on systems integrating Large Language Models (LLMs), ranging from interactive assistants to autonomous agents in