Skip to content

Implementation of the paper "Zero-shot World Models via Search in Memory", accepted at NeurIPS 2025.

Notifications You must be signed in to change notification settings

fmalato/zero_shot_world_models

Repository files navigation

Zero-shot World Models via Search in Memory

Implementation of the paper "Zero-shot World Models via Search in Memory", accepted at NeurIPS 2025.

Running

To run the code, install the requirements provided in requirements.txt with pip (or any other package manager). Then, python <file_name>.py should suffice.

Data collection

To collect data, please run python super_tux_kart_record.py. If you wish to test different environments, you can adapt the script to it, or use publicly-available datasets for it.

Important! The source code come with no data due to upload size limit. Please collect it following the instructions above before running any script.

Disclaimer

Results may vary slightly depending on the quality and amount of data used to replicate the experiments. However, overall results will remain unchanged.

About

Implementation of the paper "Zero-shot World Models via Search in Memory", accepted at NeurIPS 2025.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages