Reproduce “Anticipating Social Consequences with Distributional World Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Viet-Thanh Pham, Thuy-Trang Vu, Zhuang Li, Lizhen Qu, Gholamreza Haffari
DSWM forecasts several possible responses to each candidate social action, their likelihoods, and outcomes such as revenue or goal completion. It learns these distributions from repeated simulations and gives them to a separate actor, which chooses an action. Across five adapted environments, it has the best mean on 13 of 15 metrics; human transfer and probability calibration remain untested.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.