Rollout then Recycle: 3,718 Labelled Real-Robot Episodes from Benchmarking 7 Policies on 12 Tasks
Rollout then Recycle: 3,718 Labelled Real-Robot Episodes from Benchmarking 7 Policies on 12 Tasks
This publication releases all 3,718 episodes from the ArmnetBench v0.1 evaluation, which includes 600 teleoperated demonstrations and 3,118 rollouts of seven imitation-learning and vision-language-action policies on 12 tabletop tasks recorded on low-cost SO-101 arms. The dataset contains 73% failure rollouts, with every episode labelled by an operator as successful, suboptimal, or failure. Each episode includes three camera views, joint states, and actions, and is available on the Hugging Face Hub in LeRobot v3.0 and a reward-model format.
The authors release a dataset of 3,718 labelled episodes from the ArmnetBench v0.1 evaluation, comprising teleoperated demonstrations and policy rollouts on low-cost SO-101 arms across 12 tasks, with explicit documentation of collection methods and limitations.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- Control
- TeleoperatedOpen source S2
Reported numbers
episodes
3,718 episodes
View original evidence
arXiv:2611.03168v1 Announce Type: new Abstract: We release all 3,718 episodes from the ArmnetBench v0.1 evaluation (arXiv:2607.24481), which contains 600 teleoperated demonstrations and 3,118 rollouts of seven imitation-learning and vision-language-action policies on 12 tabletop tasks, recorded on low-cost SO-101 arms; 73% of the rollouts are failures. Every policy for a task was trained or fine-tuned on the same 50
Open source S2demonstrations
600
View original evidence
arXiv:2611.03168v1 Announce Type: new Abstract: We release all 3,718 episodes from the ArmnetBench v0.1 evaluation (arXiv:2607.24481), which contains 600 teleoperated demonstrations and 3,118 rollouts of seven imitation-learning and vision-language-action policies on 12 tabletop tasks, recorded on low-cost SO-101 arms; 73% of the rollouts are failures. Every policy for a task was trained or fine-tuned on the same 50
Open source S2rollouts
3,118
View original evidence
arXiv:2611.03168v1 Announce Type: new Abstract: We release all 3,718 episodes from the ArmnetBench v0.1 evaluation (arXiv:2607.24481), which contains 600 teleoperated demonstrations and 3,118 rollouts of seven imitation-learning and vision-language-action policies on 12 tabletop tasks, recorded on low-cost SO-101 arms; 73% of the rollouts are failures. Every policy for a task was trained or fine-tuned on the same 50
Open source S2tasks
12
View original evidence
arXiv:2611.03168v1 Announce Type: new Abstract: We release all 3,718 episodes from the ArmnetBench v0.1 evaluation (arXiv:2607.24481), which contains 600 teleoperated demonstrations and 3,118 rollouts of seven imitation-learning and vision-language-action policies on 12 tabletop tasks, recorded on low-cost SO-101 arms; 73% of the rollouts are failures. Every policy for a task was trained or fine-tuned on the same 50
Open source S2policies
7
View original evidence
arXiv:2611.03168v1 Announce Type: new Abstract: We release all 3,718 episodes from the ArmnetBench v0.1 evaluation (arXiv:2607.24481), which contains 600 teleoperated demonstrations and 3,118 rollouts of seven imitation-learning and vision-language-action policies on 12 tabletop tasks, recorded on low-cost SO-101 arms; 73% of the rollouts are failures. Every policy for a task was trained or fine-tuned on the same 50
Open source S2camera views
3
View original evidence
demonstrations and rolled out on the same cell, so the corpus contains many policies' failures on shared tasks. An operator labelled every rollout successful, suboptimal or failure. Each episode has three camera views, joint states and actions and is available on the Hugging Face Hub in LeRobot v3.0 and a reward-model format. This note documents how the episodes were collected and the limitations that could affect th
Open source S3robots
1
View original evidence
arXiv:2611.03168v1 Announce Type: new Abstract: We release all 3,718 episodes from the ArmnetBench v0.1 evaluation (arXiv:2607.24481), which contains 600 teleoperated demonstrations and 3,118 rollouts of seven imitation-learning and vision-language-action policies on 12 tabletop tasks, recorded on low-cost SO-101 arms; 73% of the rollouts are failures. Every policy for a task was trained or fine-tuned on the same 50
Open source S2
Dataset details reported by the source
- Version
- LeRobot v3.0 Open source S3
- Modalities
- camera views Open source S3joint states Open source S3
- Tasks
- 12 tabletop tasks Open source S2
- Embodiments
- SO-101 arms Open source S2
- dataset: not_reported
- paper: not_reported
- code: not_reported
- weights: not_reported
What remains unknown
Dataset licensing and commercial-use rights are not established here.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
arXiv:2611.03168v1 Announce Type: new Abstract: We release all 3,718 episodes from the ArmnetBench v0.1 evaluation (arXiv:2607.24481), which contains 600 teleoperated demonstrations and 3,118 rollouts of seven imitation-learning and vision-language-action policies on 12 tabletop tasks, recorded on low-cost SO-101 arms; 73% of the rollouts are failures. Every policy for a task was trained or fine-tuned on the same 50
Open source S2
demonstrations and rolled out on the same cell, so the corpus contains many policies' failures on shared tasks. An operator labelled every rollout successful, suboptimal or failure. Each episode has three camera views, joint states and actions and is available on the Hugging Face Hub in LeRobot v3.0 and a reward-model format. This note documents how the episodes were collected and the limitations that could affect th
Open source S3
Implications for data suppliers
RoboSignal interpretation and collection questions, not statements of buyer demand.
- Confirm the required data type and collection setting with the buyer; this source does not establish a complete collection specification.
- Compare the reported units and scope before using these quantities in a budget. Recording hours, sensor-hours and trajectories are different measures.
- Check the exact dataset version, license, modalities and task coverage against the intended use.
- Validate demand and acceptance criteria with a buyer before scaling. Publication, popularity and a research result do not establish a purchase commitment.
Source：arXiv Robotics — research abstracts · arxiv.org