Angela Radulescu

This study introduces a behavioral paradigm to measure reward generalization and links broader generalization to positive overgeneralization and subclinical bipolar traits.

The paper shows prefrontal circuits coordinate to encode attention-weighted value signals during learning, shaping state representations in complex environments.

This study uses recurrent neural networks to infer human attention dynamics during learning, finding that attention reflects an interaction between value-based learning and hypothesis testing.

This paper introduces MAIRL, a multi-agent inverse reinforcement learning framework that decomposes joint value functions into individual value maps and interaction terms, revealing interpretable social value representations in mice and primates.

This study models human eye movements during immersive visual search using a reinforcement learning framework, finding that model predictions align with human gaze behavior.