Wmbench

WmbenchWmbench

Wmbench

Stealth modeStealth Mode

A benchmark for evaluating video world models as surrogate policy evaluators for embodied robot policies.

DR 0Speed 902026Active
Rate

About

WMBench is a real-robot benchmark constructed from teleoperation data and matched policy rollouts. It enables controlled comparisons across model families, action encodings, rollout horizons, and evaluation metrics for video world models, particularly in their ability to serve as reliable surrogate evaluators for embodied robot policies. The research highlights that evaluator quality is driven by long-horizon, action-faithful rollout consistency rather than short-term visual realism.
Frequently asked

What does Wmbench do?

WMBench is a real-robot benchmark constructed from teleoperation data and matched policy rollouts. It enables controlled comparisons across model families, action encodings, rollout horizons, and evaluation metrics for video world models, particularly in their ability to serve as reliable surrogate evaluators for embodied robot policies. The research highlights that evaluator quality is driven by long-horizon, action-faithful rollout consistency rather than short-term visual realism.

Is Wmbench trustworthy and reputable?

StartupHub's Data Trust & Reputation score for Wmbench is 26 out of 100, based on site security posture and privacy practices.

When was Wmbench founded?

Wmbench was founded in 2026.

What industry does Wmbench operate in?

Wmbench operates in Robotics, World Model, AI Testing, Benchmark, Embodied AI, Reinforcement Learning.

New entrants in Robotics
Last 90 days
43
New entrants
14.3
Per month
3.3
Per week

An entrant is a company tagged Robotics whose domain was first registered in the window, counted from registry records in the StartupHub directory. 3 of them registered in the last 30 days. Registry detection runs two to three weeks behind registration, so recent weeks are a floor.

Comments

No comments yet. Be the first to share your take.