Add model card
#1
by nielsr HF Staff - opened
README.md
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
pipeline_tag: robotics
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
# ACG-WAM: World-Action Modeling via Action-Conditioned Geometric Latent Prediction
|
| 7 |
+
|
| 8 |
+
This repository contains the **ACG-WAM-RoboTwin-40K** policy presented in [ACG-WAM: World-Action Modeling via Action-Conditioned Geometric Latent Prediction](https://huggingface.co/papers/2610.06965).
|
| 9 |
+
|
| 10 |
+
ACG-WAM is a world-action model that learns to predict the geometric consequences of actions through an auxiliary Action-Conditioned Geometric Joint-Embedding Predictive Architecture (ACG-JEPA). The released checkpoint is the Joint ACG-JEPA policy trained on clean and randomized RoboTwin 2.0 demonstrations for 40,000 optimizer updates.
|
| 11 |
+
|
| 12 |
+
- Project page: [https://RoboOpus.github.io/ACG-WAM](https://RoboOpus.github.io/ACG-WAM)
|
| 13 |
+
- Code: [https://github.com/RoboOpus/ACG-WAM](https://github.com/RoboOpus/ACG-WAM)
|
| 14 |
+
|
| 15 |
+
See the code repository for installation, training, and evaluation instructions.
|