Reproduce “PSM-WSM: Predictively Sufficient Multimodal World Social Modeling” 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

Chunlei Meng, Zhenyu Yu, Pengbin Feng, Jacqueline J. Pang, Jiabin Luo, Chih-Ting Liao, Zhaolu Kang, Chun Ouyang and 1 more

PSM-WM constructs multimodal states to predict interactions evolve, rather than only describe the present. It encodes language, audio, and video histories, trains states to predict future representations, and models event-conditioned transitions. On Hi-EF, it reports 75.76 weighted F1, compared with 74.91 for ModularAgent. Controlled MELD probes favor its state construction, but provide empirical evidence, not a guarantee of predictive sufficiency.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.