Official project page for Improving Policy Learning via Language-Guided State Representation in World Models, accepted by IEEE Robotics and Automation Letters in July 2026.
LGSR uses language instructions to guide compact, task-relevant state representations in world models for long-horizon robotic policy learning. The project page presents the method, CALVIN and LIBERO-Long results, representation analyses, and additional experiments.
The site is static and has no build step. From the repository root, run:
python3 -m http.server 8000Then open http://127.0.0.1:8000/.
index.html: page content and metadata.static/css/: page and framework styles.static/js/: navigation and citation interactions.static/images/: paper figures, additional visualizations, and favicons.static/pdfs/: accepted manuscript.
The repository is published directly with GitHub Pages. All runtime assets are stored locally, so deployment does not require package installation or a build command.
The presentation is inspired by the Nerfies and VGGT project pages and is adapted from the Academic Project Page Template.
The project-page code is distributed under the Creative Commons Attribution-ShareAlike 4.0 International License.