Reproduce “PrismWAM: Decoupling State Representations in World Action 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
Tianyu Xin, Wanqi Zhong, Bowen Hu, Zhangchi Guo, Lijie Wang, Yiming Wang, Haoran Liao, Heyang Yu and 3 more
PrismWAM separates semantic context from physical state in a world action model. It uses RGB to guide task interpretation, depth to predict physical changes, and actions are generated from that rollout. On three real-robot tasks, pooled out-of-distribution success was 68.6% across 35 trials, versus 48.6% for OpenWAM; this small comparison does not isolate representation as the cause.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.