OnlyAct checkpoints

Weights for OnlyAct: Extracting Action Representations from Foundation Models for World Models (NeurIPS 2026).

• Code, usage and the evaluation of every table: https://github.com/runjiali-rl/onlyact
• Paper: https://arxiv.org/abs/2611.14420
• Project page: https://runjiali-rl.github.io/onlyact/

This repository was called liguang0115/justact (the method's earlier name, JustAct); the old address redirects here. On 2026-11-11 the files named after JustAct were renamed to OnlyAct. Their contents are unchanged and SHA256SUMS lists the same checksums.

Download

# in the root of the GitHub repository (README.md is excluded so that checkpoints/README.md is not overwritten)
hf download liguang0115/onlyact --local-dir checkpoints --exclude README.md .gitattributes
cd checkpoints && sha256sum -c SHA256SUMS

Files

file	size	content	used by
onlyact/adapter_config.json, onlyact/adapter_model.safetensors	34 MB	OnlyAct LoRA adapters for V-JEPA 2 ViT-L (PEFT format)	everything
onlyact/projection.safetensors	2.6 MB	the fixed random 1024 -> 640 projection (weight, bias)	everything
adaworld_lam_retrained.safetensors	1.9 GB	AdaWorld latent action model trained on our frame pairs	Tables 1, 2, 4
dreamdojo_lam_retrained.safetensors	2.8 GB	DreamDojo latent action model trained on our frame pairs	Tables 1, 2, 4
worldmodel/{no_action,adaworld_lam,dreamdojo_lam,onlyact}.safetensors	223-246 MB each	trained weights (LoRA + action conditioning) of the four AdaWorld world models	Tables 2, 3
worldmodel/onlyact_whitening.safetensors	3.3 MB	whitening statistics of the OnlyAct input of the world model	Tables 2, 3
downstream/idm_{dmlab,minecraft,habitat}.safetensors, downstream/readout_nuscenes.safetensors	about 1 MB each	inverse-dynamics classifiers (Follow metric) and the nuScenes forward-motion readout	Table 3
downstream/step0/*.safetensors, downstream/standardise/*.safetensors	small	action interfaces before fine-tuning and their standardisation statistics	Table 3

Not included, downloaded from their authors: V-JEPA 2 ViT-L (facebook/vjepa2-vitl-fpc64-256, MIT licence), the AdaWorld world model (Little-Podi/AdaWorld, Apache License 2.0) and the OpenCLIP ViT-H-14 image encoder. See checkpoints/README.md in the GitHub repository for where each one goes.

Licence

These weights are released under the Apache License 2.0. The world-model weights are adapters on the AdaWorld world model, which is itself fine-tuned from Stable Video Diffusion, so Stability AI's terms for that model also apply. The training and evaluation datasets keep their own licences, and several of them are for non-commercial use only (listed in the paper's appendix and in the GitHub repository).

Citation

@article{li2026onlyact,
  title   = {OnlyAct: Extracting Action Representations from Foundation Models for World Models},
  author  = {Li, Runjia and Chen, Minghao and Xie, Junyu and Torr, Philip and Vedaldi, Andrea and Jakab, Tomas},
  journal = {arXiv preprint arXiv:2611.14420},
  year    = {2026}
}