WorldFoundry catalogs 270 models for world-model evaluation
Original title: WorldFoundry: Towards Unified Infrastructure for World Intelligence
The open-source stack also lists 74 benchmarks and 62 metrics, separating execution evidence from scores.
A paper defines a generative world model as a system that keeps persistent state, updates it from actions, and produces coherent trajectories under observed and counterfactual interventions. It introduces WorldFoundry, open-source infrastructure for reproducible inference and evaluation across video, 3D and 4D, embodied action, and interactive systems. Manifests and evidence-aware scorecards are intended to tie claims to artifacts.
- The catalog has 270 model entries, 74 benchmarks, and 62 registered metrics.
- Readiness and execution evidence are recorded separately from evaluative claims.
- Modality adapters preserve native inputs, execution semantics, and outputs.
- The framework spans grounding, state, control, prediction, and closed-loop interaction.
From the original article
1 min read at huggingface.co
Visual generative models increasingly blur the line between synthesizing observations and simulating environments. Yet visual plausibility alone does not constitute successful world modeling: compelling outputs may lack persistent state, physical dynamics, or causally consistent responses to interactions.
By Juanxi Tian, Jiadong Hou, Baiang Li · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.