Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,113 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
# GigaBrain Challenge 2026 – Task 7 VLA Policy Model
|
| 6 |
+
|
| 7 |
+
This repository provides the **Vision-Language-Action (VLA) policy model for Task 7** of the **GigaBrain Challenge 2026**.
|
| 8 |
+
|
| 9 |
+
Official Challenge Website:
|
| 10 |
+
|
| 11 |
+
https://gigaai-research.github.io/GigaBrain-Challenge-2026/
|
| 12 |
+
|
| 13 |
+
The GigaBrain Challenge aims to advance research in **embodied intelligence**, including:
|
| 14 |
+
|
| 15 |
+
- Vision-Language-Action (VLA) models
|
| 16 |
+
- world models
|
| 17 |
+
- robot learning systems
|
| 18 |
+
- closed-loop embodied agents
|
| 19 |
+
|
| 20 |
+
---
|
| 21 |
+
|
| 22 |
+
# Task Overview
|
| 23 |
+
|
| 24 |
+
**Task 7** is one of the official tasks in the GigaBrain Challenge 2026 benchmark suite.
|
| 25 |
+
|
| 26 |
+
Each task evaluates an embodied agent’s ability to generate actions based on multimodal observations and task instructions.
|
| 27 |
+
|
| 28 |
+
For full task descriptions please refer to:
|
| 29 |
+
|
| 30 |
+
https://gigaai-research.github.io/GigaBrain-Challenge-2026/index.html#tracks
|
| 31 |
+
|
| 32 |
+
---
|
| 33 |
+
|
| 34 |
+
# Model Description
|
| 35 |
+
|
| 36 |
+
This repository contains the **VLA policy model used for Task 7**.
|
| 37 |
+
|
| 38 |
+
The model is designed to:
|
| 39 |
+
|
| 40 |
+
- take **visual observations and language instructions** as input
|
| 41 |
+
- generate **robot actions**
|
| 42 |
+
- interact with a **world model** in a closed-loop evaluation environment
|
| 43 |
+
|
| 44 |
+
The VLA model serves as the **action generation component** in the challenge evaluation pipeline.
|
| 45 |
+
|
| 46 |
+
During evaluation:
|
| 47 |
+
Observation → VLA Model → Action → World Model → Next Observation
|
| 48 |
+
|
| 49 |
+
This loop continues until the task is completed or the episode terminates.
|
| 50 |
+
|
| 51 |
+
Participants may use this model as:
|
| 52 |
+
|
| 53 |
+
- a reference implementation
|
| 54 |
+
- a starting point for model development
|
| 55 |
+
- a component in their own embodied AI systems
|
| 56 |
+
|
| 57 |
+
---
|
| 58 |
+
|
| 59 |
+
# Dataset
|
| 60 |
+
|
| 61 |
+
The dataset for all tasks is provided in the official challenge dataset repository:
|
| 62 |
+
|
| 63 |
+
https://huggingface.co/datasets/open-gigaai/CVPR-2026-WM-Track-Dataset
|
| 64 |
+
|
| 65 |
+
Access requires:
|
| 66 |
+
|
| 67 |
+
- Team registration
|
| 68 |
+
- Agreement to the challenge license
|
| 69 |
+
|
| 70 |
+
Please request access on the dataset page.
|
| 71 |
+
|
| 72 |
+
---
|
| 73 |
+
|
| 74 |
+
# License and Usage Terms
|
| 75 |
+
|
| 76 |
+
This model is released as part of the GigaBrain Challenge 2026.
|
| 77 |
+
|
| 78 |
+
Use of this model and the associated datasets is governed by the GigaBrain Challenge 2026 Data & Model License Agreement.
|
| 79 |
+
|
| 80 |
+
Key terms include:
|
| 81 |
+
|
| 82 |
+
Intended for research and competition participation
|
| 83 |
+
|
| 84 |
+
Dataset redistribution is not permitted
|
| 85 |
+
|
| 86 |
+
Commercial use requires permission from the organizers
|
| 87 |
+
|
| 88 |
+
For full license terms please refer to the dataset repository:
|
| 89 |
+
|
| 90 |
+
https://huggingface.co/datasets/open-gigaai/CVPR-2026-WM-Track-Dataset
|
| 91 |
+
|
| 92 |
+
# Citation
|
| 93 |
+
|
| 94 |
+
If you use this model or dataset in academic research, please cite the challenge:
|
| 95 |
+
```bibtex
|
| 96 |
+
@misc{gigabrain_challenge_2026,
|
| 97 |
+
title={GigaBrain Challenge 2026},
|
| 98 |
+
year={2026},
|
| 99 |
+
howpublished={https://gigaai-research.github.io/GigaBrain-Challenge-2026/}
|
| 100 |
+
}
|
| 101 |
+
```
|
| 102 |
+
|
| 103 |
+
# Disclaimer
|
| 104 |
+
|
| 105 |
+
This model is provided as-is without warranty of any kind.
|
| 106 |
+
|
| 107 |
+
The organizers are not responsible for any issues arising from its use.
|
| 108 |
+
|
| 109 |
+
# Contact
|
| 110 |
+
|
| 111 |
+
For questions about the challenge, please visit the official website:
|
| 112 |
+
|
| 113 |
+
https://gigaai-research.github.io/GigaBrain-Challenge-2026/
|