HumanVerse-500 Dataset and λ₀ Policy for Humanoid Loco-Manipulation
Overview
A 500-hour dataset and λ₀ policy trained on egocentric human data aim to enhance humanoid robot loco-manipulation. New research introduces HumanoidTTT for efficient control reuse.
Generated from attributed reports · 2 days agoUpdated
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Artifact availability · code: reported pending
“Code: https://github.com/AIGeeksGroup/HumanoidTTT. Website: https://aigeeksgroup.github.io/HumanoidTTT.”
Exact source · revision 1Source owner not reported
Reported quantity · hours: 500 hours · unique elapsed hours
“a 500-hour dataset of diverse human loco-manipulation behaviors in open-world environments”
Exact source · revision 1Source owner not reported
Developments
- 2026-10-02 04:00 UTC · 1 reportsTowards a General Humanoid Loco-Manipulation Model via EgocearXiv Robotics — research abstracts：Towards a General Humanoid Loco-Manipulation Model via Egocentric Whole-Body Human Data Pretraining
- 2026-10-02 04:00 UTC · 1 reportsHumanoidTTT: Test-Time Capability Reuse for Efficient HumanoarXiv Robotics — research abstracts：HumanoidTTT: Test-Time Capability Reuse for Efficient Humanoid Control
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsHumanoidTTT: Test-Time Capability Reuse for Efficient Humanoid Control
This paper introduces HumanoidTTT, a framework for test-time capability reuse in continual humanoid control. It enables reliable and efficient reuse of validated motion capabilities through Selective Full-Motion Reuse and Test-Time Capability Consolidation, achieving a 16.4× end-to-end speedup over fresh generation.
- arXiv Robotics — research abstractsTowards a General Humanoid Loco-Manipulation Model via Egocentric Whole-Body Human Data Pretraining
The paper presents HumanVerse-500, a 500-hour dataset of human loco-manipulation behaviors collected with a lightweight wearable system. It introduces λ₀, a whole-body humanoid vision-language-action policy trained through three stages, achieving state-of-the-art performance on real-world tasks and analyzing how human data supports downstream control.
Event attention history
Current attention 3·Peak within the comparable range 9（2026-10-02 06:00 UTC）·Change within the comparable range over 24 hours -49%
The trend compares participants observed continuously; its scope may be narrower than current attention totals. Hover or click to explore hourly observations, or use the left and right arrow keys.