Buyer guide

World-model datasets built around transitions

Source or commission world model training data that connects observations, actions, state, and outcomes through time.

Synthetic schema sampleObservation → action → next state
01
00.000sObservergb.player_view
02
00.050sActmove.forward + camera.delta
03
00.100sTransitionpose.x +0.84 m
04
00.150sOutcomeprogress 0.31 → 0.36

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

Direct answer

World model training data is a collection of temporally ordered episodes used to learn how an environment changes. For controllable models, each episode should align observations with the actions, state transitions, and outcomes that produced the next moment.

The episode is the useful unit

Loose clips can teach visual priors. An episode preserves the initial condition, action history, environment state, task context, and terminal result needed to learn dynamics.

Each stream should resolve to one declared clock. Sampling rate, dropped frames, action delay, and uncertainty belong in the release record, not in private implementation notes.

  • Observed pixels, audio, depth, language, or sensors
  • Raw and normalized action signals
  • Measured engine or simulator state
  • Events, progress, reward, failure, and terminal state

Hours do not describe coverage

Nominal duration can hide repetition. A useful release reports coverage across environments, task states, action combinations, outcomes, rare events, failures, and recoveries.

Define the coverage matrix before collection. It becomes the contract connecting the research target to collection quotas and release acceptance.

Rights are part of the dataset

Download access is not commercial permission. The release needs a traceable source, rights holder, allowed uses, redistribution terms, attribution requirements, restrictions, and withdrawal path.

ActionTrajectories keeps external research references separate from datasets commissioned or supplied under a commercial agreement.

Questions this page should settle

What data does a world model need?

It needs ordered evidence of environment change. Controllable systems benefit from synchronized observations, actions, state, events, and outcomes.

Is video alone a world-model dataset?

Video can support predictive pretraining, but it does not directly identify which action caused a transition. Action and state streams make control and evaluation more explicit.

Can ActionTrajectories source custom data?

Yes. A project begins with the target environment, action space, required state, evaluation plan, scale, rights, and delivery constraints.

Bring us the training target.

We will map it to the observations, actions, state, outcomes, rights, and delivery evidence the dataset must contain.

Request access