Excited to see this out after a while (evident in the LLMs tested!). Grateful for the support of my supervisors @lei-zhang.bsky.social and @thepsychologist.bsky.social, as well as @arkadykonovalov.bsky.social and Xintong 'Janice' Zhong for help with data collection and modeling!
📣 New preprint 📣 An exciting collaboration with Lei Zhang's lab spearheaded by Aamir Sohail. We use game tasks and models typically used to estimate mentalization ability in humans, and we test them across a variety of commercially available LLMs. arxiv.org/abs/2608.26291