Skip to content
Hugging Face Daily Papers·· 2026-10-19AI score45

Stateful Harness: A world model framework that separates rule execution from video rendering

State World Models: Execute the Rules, Generate the Pixels

AI briefing

Researchers propose the Stateful Harness framework, which separates world state evolution from video rendering: rules maintain the world state, a frozen video model renders each event, and the generated clips are then used to update the state. The team built WorldState-Bench, containing 50 pairs of counterfactual twin samples; with rule prompts alone, the best of 11 video backends was only 32% correct; MiniMax-H3 rose to 64% when the ending was explicitly specified, and reached 92% with the full framework connected.

Open original arxiv.org