Home-Robot-Arm-Demo-Efficiency-Dataset
Simulated at-home robot-arm learning curves (success rate vs. number of self-collected teleop demonstrations) for 480 configurations across 5 home-manipulation tasks, 5 learning methods (behavior cloning from scratch, ACT/ALOHA, Diffusion Policy, OpenVLA fine-tuning, OpenVLA co-training), and realistic hobbyist conditions, grounded in the 2023-2024 low-cost/personal robot-learning literature. Forward-looking task predicts whether a task will succeed reliably by 100 demos from just a 10-demo pilo
About Dataset
Scoped for a solo hobbyist on a low-cost desk arm, not a lab or company exactly the audience ACT/ALOHA, Mobile ALOHA, and OpenVLA explicitly target. Simulates the learning curve a solo operator actually sees across 5 tasks, 5 methods, and 3 hobbyist-controllable conditions (co-training, session consistency, setup control). Ten sources (2021–2024) documented in demo_efficiency_theory.py, including ACT/ALOHA, Diffusion Policy, Mobile ALOHA, OpenVLA, Open X-Embodiment, BridgeData V2, DROID, RT-1, robomimic, and Behavior Transformers.
Purpose of Dataset
Lets A Hobbyist Decide, From A Small Saturday-morning Pilot Batch Of Demos, Whether To Keep Collecting Toward 100 Or Switch Methods Without Guessing.
Dataset Metadata
Attribution 4.0 International (CC BY- 4.0)
India
Science, Technology and Research
Jai Mali
Structured
NA
NA
N.A.
08/08/26 16:32:58
Open
Hosted
Primary
Activity Overview
0
0 
1.22 MB 
4
Tags
- Robotics
License Control
Attribution 4.0 International (CC BY- 4.0)
Home-Robot-Arm-Demo-Efficiency-Dataset.json ( 1.22 MB )
To preview this file, you need to be a registered user. Please complete the registration process to gain access and continue viewing the content.
Data Quality Score Beta
Version Control
Version 1(1.22 MB)
- JAI MALI·2 day(s) ago
Home-Robot-Arm-Demo-Efficiency-Dataset.json
© 2026 - Copyright AIKosh. All rights reserved.