Add paper link, GitHub repository, and improve metadata

#2
by nielsr HF Staff - opened
Files changed (1) hide show
  1. README.md +20 -30
README.md CHANGED
@@ -1,8 +1,9 @@
1
  ---
2
- task_categories:
3
- - robotics
4
  language:
5
  - en
 
 
 
6
  tags:
7
  - RDT
8
  - rdt
@@ -10,14 +11,18 @@ tags:
10
  - manipulation
11
  - bimanual
12
  - ur5e
13
- - webdatset
14
  - vision-language-action
15
- license: apache-2.0
16
  ---
17
 
 
 
 
 
18
  ## Dataset Summary
19
 
20
- This dataset provides shards in the **WebDataset** format for fine-tuning [RDT-2](https://rdt-robotics.github.io/rdt2/) or other policy models on **bimanual manipulation**.
21
  Each sample packs:
22
 
23
  * a **binocular RGB image** (left + right wrist cameras concatenated horizontally)
@@ -91,7 +96,7 @@ shard-000002.tar
91
 
92
  ### 1) Official Guidelines to fine-tune RDT 2 series
93
 
94
- Use the example [scripts](https://github.com/thu-ml/RDT2/blob/cf71b69927f726426c928293e37c63c4881b0165/data/utils.py#L48) and [guidelines](https://github.com/thu-ml/RDT2/blob/cf71b69927f726426c928293e37c63c4881b0165/data/utils.py#L48):
95
 
96
  ### 2) Minimal Loading example
97
 
@@ -138,9 +143,9 @@ def get_train_dataset(shards_dir):
138
 
139
  return dataset
140
 
141
- with open(os.path.join("<Dataset Diretory>", "instructions.json") as fp:
142
  instructions = json.load(fp)
143
- dataset = get_train_dataset(os.path.join("<Dataset Diretory>", "shards"))
144
  ```
145
  ---
146
 
@@ -153,15 +158,12 @@ dataset = get_train_dataset(os.path.join("<Dataset Diretory>", "shards"))
153
 
154
  ## Citation
155
 
156
- If you use this dataset, please cite the dataset and your project appropriately. For example:
157
-
158
  ```bibtex
159
- @software{rdt2,
160
- title={RDT2: Enabling Zero-Shot Cross-Embodiment Generalization by Scaling Up UMI Data},
161
- author={RDT Team},
162
- url={https://github.com/thu-ml/RDT2},
163
- month={September},
164
- year={2025}
165
  }
166
  ```
167
 
@@ -169,17 +171,5 @@ If you use this dataset, please cite the dataset and your project appropriately.
169
 
170
  ## License
171
 
172
- * **Dataset license:** Apache-2.0 (unless otherwise noted by the maintainers of your fork/release).
173
- * Ensure compliance when redistributing derived data or models.
174
-
175
- ---
176
-
177
- ## Maintainers & Contributions
178
-
179
- We welcome fixes and improvements to the conversion scripts and docs (see https://github.com/thu-ml/RDT2/tree/main#troubleshooting).
180
- Please open issues/PRs with:
181
-
182
- * OS + Python versions
183
- * Minimal repro code
184
- * Error tracebacks
185
- * Any other helpful context
 
1
  ---
 
 
2
  language:
3
  - en
4
+ license: apache-2.0
5
+ task_categories:
6
+ - robotics
7
  tags:
8
  - RDT
9
  - rdt
 
11
  - manipulation
12
  - bimanual
13
  - ur5e
14
+ - webdataset
15
  - vision-language-action
16
+ arxiv: 2602.03310
17
  ---
18
 
19
+ # RDT2 Dataset
20
+
21
+ [Project page](https://rdt-robotics.github.io/rdt2/) | [Paper](https://huggingface.co/papers/2602.03310) | [GitHub](https://github.com/thu-ml/RDT2)
22
+
23
  ## Dataset Summary
24
 
25
+ This dataset provides shards in the **WebDataset** format for fine-tuning [RDT-2](https://rdt-robotics.github.io/rdt2/) or other policy models on **bimanual manipulation**.
26
  Each sample packs:
27
 
28
  * a **binocular RGB image** (left + right wrist cameras concatenated horizontally)
 
96
 
97
  ### 1) Official Guidelines to fine-tune RDT 2 series
98
 
99
+ Use the official [scripts](https://github.com/thu-ml/RDT2) and [guidelines](https://github.com/thu-ml/RDT2) provided in the GitHub repository.
100
 
101
  ### 2) Minimal Loading example
102
 
 
143
 
144
  return dataset
145
 
146
+ with open(os.path.join("<Dataset Directory>", "instructions.json")) as fp:
147
  instructions = json.load(fp)
148
+ dataset = get_train_dataset(os.path.join("<Dataset Directory>", "shards"))
149
  ```
150
  ---
151
 
 
158
 
159
  ## Citation
160
 
 
 
161
  ```bibtex
162
+ @article{rdt2,
163
+ title={RDT2: Exploring the Scaling Limit of UMI Data Towards Zero-Shot Cross-Embodiment Generalization},
164
+ author={RDT Team},
165
+ journal={arXiv preprint arXiv:2602.03310},
166
+ year={2026}
 
167
  }
168
  ```
169
 
 
171
 
172
  ## License
173
 
174
+ * **Dataset license:** Apache-2.0.
175
+ * Ensure compliance when redistributing derived data or models.