CauSyncVLA / README.md
SSSSphinx's picture
Link the CauSyncVLA paper
cf192de verified
|
Raw History Blame Contribute Delete
1.04 kB
---
tags:
- arxiv:2611.25667
- robotics
- vision-language-action
- safetensors
---
# CauSyncVLA LIBERO updater
The LIBERO KV-cache updater from [CauSyncVLA: Causal Latent World Modeling for Asynchronous Robot Control](https://arxiv.org/abs/2611.25667) conditions a frozen OpenPI π0.5-LIBERO expert on the latest published visual cache and executed state/action history.
| File | Training step | History |
| --- | ---: | ---: |
| `libero/updater.safetensors` | 12,000 | 48 simulator steps |
The safetensors file contains the learned parameters and original KV/state normalization buffers. Its metadata records the architecture and expected action-normalization identity. The export was checked against every tensor of the training checkpoint.
Download the matching frozen expert through [OpenPI](https://github.com/Physical-Intelligence/openpi). [CauSyncVLA code and evaluation instructions](https://github.com/TwSphinx54/CauSyncVLA) describe installation, checkpoint conversion and reproduction of the LIBERO evaluation.