Embodied AI

Central hub for Embodied AI experiments within the Robotics Security Division: foundation model evaluation, adversarial robustness, and generalization research across robot policies and simulation environments.

Experiments

ID	Title	Status
Experiment 1	Test scene overfitting in robot foundation models	In progress
Experiment 2	Latent representation geometry at the S2→S1 interface of GR00T N1.6	First result
Experiment 3	Adversarial visual patches on generalist robot policies	In progress
Experiment 4	Boundary task steering on frozen GR00T policies	Current result

Structure

Each experiment lives in its own folder (experiment_N/) with a README describing the hypothesis, setup, and results. Sub-experiments (e.g., per-model runs) are nested as experiment_N/N.M/.

Contributing

To add a new experiment:

1. Create experiment_N/ with a README.md and an index.html (the GitHub Pages entry point).
2. Add sub-experiments as experiment_N/N.M/ if needed.
3. Register the experiment in the table above and on the landing page.

A starting point is available in _template/ — it mirrors the structure of Experiment 1. Feel free to ignore it and build your results page however works best for your data.