Technology
Applications
About us
NewsContact us
← Back to the newsroom
Newsroom

Awomo Achieves #1 Rankings Across Four Robotics Benchmarks

Research Release
#1 across all four benchmarks as of 1 October 2026: RoboDojo 35.3 first among all WAM and VLA, MolmoSpaces 64.5% first among all WAM and VLA, RoboTwin 2.0 96.1% first overall, LIBERO-Plus 93.8% first overall

Awomo leads WAM and VLA entries on RoboDojo and MolmoSpaces, and ranks first overall on RoboTwin 2.0 and LIBERO-Plus.

As of October 1, 2026, Awomo achieved first-place rankings across four robotics benchmarks, demonstrating performance in manipulation, bimanual coordination, and adaptation to changing environments.

  • RoboDojo: 35.3 — #1 among WAM and VLA entries
  • MolmoSpaces: 64.5% — #1 among WAM and VLA entries
  • RoboTwin 2.0: 96.1% — #1 overall
  • LIBERO-Plus: 93.8 — #1 overall

Performance Across Four Benchmarks

RoboDojo evaluates robotic capabilities across five dimensions: generalization, precision, long-horizon task execution, memory, and open-ended tasks. Awomo leads WAM and VLA entries with an average score of 35.3. The October 2 leaderboard lists Awomo-0.5 at 35.34, retaining this position.

MolmoSpaces evaluates object interactions, including opening, closing, picking, and placing. Awomo achieves a Combined Average of 64.5%, ranking first among WAM and VLA entries.

RoboTwin 2.0 evaluates bimanual manipulation across standard and randomized conditions. Awomo ranks first overall with an average of 96.1%, scoring 96.04% in the standard setting and 96.20% in the randomized setting.

LIBERO-Plus evaluates robustness across seven categories of variation: camera viewpoints, instruction wording, scene layouts, backgrounds, noise, lighting, and robot initial states. Awomo achieves an average of 93.8%, ranking first overall and exceeding the previous best result of 92.4% by 1.4 percentage points.

Building Reusable Robotic Capabilities

Awomo develops reusable “atomic capabilities”: understanding instructions, assessing spatial relationships, executing precise actions, and using past observations. These foundational abilities support different combinations of tasks and environments.

The results reflect Awomo’s focus on strengthening these capabilities and translating improvements into measurable robotic performance. They support our broader goal of enabling machines to understand the physical world and act reliably within it.

Explore the research and results: