Research library

World-model data, translated into requirements

Practical guides for research, engineering, data partnerships, procurement, and rights review.

Synthetic schema sampleObservation → action → next state

00.000sObservergb.player_view

00.050sActmove.forward + camera.delta

00.100sTransitionpose.x +0.84 m

00.150sOutcomeprogress 0.31 → 0.36

The values are synthetic and illustrate the public specification. They are not measurements from a commercial dataset.

World-model, robotics, and embodied-AI research teams

Dyna-2's Million-Hour World-Action Model

What Dyna Robotics' million-hour human-video result shows about cross-embodiment transfer, world modeling, and the next robotics data bottleneck.

Research leads planning a new world-model data program

A practical guide to observations, actions, state, outcomes, time alignment, and coverage for world-model training datasets.

Teams deciding whether video-only pretraining is enough

How action labels change what a world model can learn, and when large passive-video corpora still provide value.

Labs sourcing or collecting action-conditioned gameplay

A practical evaluation guide for gameplay datasets, from control capture and engine state to rights, replay filtering, and release evidence.

Data buyers, research engineers, and procurement reviewers

A buyer-side evaluation framework covering alignment, coverage, state validity, leakage, rights, and loader-level QA.

Data engineers instrumenting a gameplay collection pipeline

A field-level guide to recording game state, controls, events, versions, and coordinate systems for machine learning.

Procurement, legal, data partnerships, and research leadership

How open research datasets differ from commercially licensed training data, and why download access is not resale permission.

Dataset architects and research infrastructure teams

The minimum episode, stream, timebase, provenance, rights, and quality fields needed for a portable trajectory release.

Technical buyers comparing emerging model and dataset terminology

A clear distinction between predictive environment models and systems that jointly represent, predict, or generate actions.