what is zero-wam fr?

Zero-WAM is a research project whose README is currently a placeholder rather than a working codebase. The repository's short description calls it an in-context world model aimed at zero-shot robotic task generalization, and its topic tags mention video-action models, world-action models, and zero-shot generalization. In plain terms, this points toward research on teaching a robot, or a model that controls a robot, to handle new tasks it has not been specifically trained on, by having it build an internal sense of how the world behaves from context rather than from task-specific training alone. The README itself gives almost no further detail. It states the project's title and tagline, links to a separate project page hosted outside the repository, and says the dataset and trained models will be released as open source at a later date, asking visitors to check back. There is no installation guide, no code walkthrough, no example usage, and no description of the model architecture or training data within the README as it currently stands. Given this, the repository should be understood as an early announcement or placeholder tied to a research paper or project page, rather than a usable tool or library at this time. Anyone interested in the technical details would need to visit the linked project page, since the README does not contain that information itself. Stars on the repository likely reflect interest generated by the announcement or an associated publication rather than usage of any released code, since none appears to be available yet based on the README content.