Menu

Technical glossary

Use the same words for the same data

Concise definitions for the model, collection, labeling, release, and evaluation concepts behind action trajectories.

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.

W

World model

A world model is a learned representation of how an environment changes over time. Depending on its design, it can predict future observations, rewards, terminal conditions, or compact latent states. An agent can use those predictions to compare possible futures before acting. The label describes a function, not one architecture: a model can operate in pixels or latent space, use deterministic or probabilistic predictions, and receive actions as inputs or learn only from passive sequences.

A

Action-conditioned world model

An action-conditioned world model predicts future observations, state, events, or rewards using an action, action sequence, or policy command as an explicit input. Conditioning matters because one scene can lead to different valid futures when the agent turns, waits, accelerates, selects a tool, or changes a control target. The model estimates what can happen under a specified intervention instead of predicting only the most likely continuation of a passive sequence.

W

World action model

World action model is an emerging term for a system that makes environment dynamics and action part of the same learning problem. Depending on the implementation, actions can condition a predicted future, be predicted jointly with video or state, or act as policy outputs. The phrase does not define one standard architecture, control level, or dataset schema, so a model description still needs to state what it observes, what it predicts, and how action enters the interface.

T

Trajectory

An ordered sequence of observations, actions, states, and optional rewards or outcomes belonging to one agent or episode.

E

Episode

A bounded attempt or interaction sequence with a defined beginning, context, and terminal condition.

O

Observation

Information available to an agent or model at a point in time, such as pixels, audio, depth, language, or sensor measurements.

A

Action

A control input or decision intended to influence the environment, recorded at a defined semantic level and time.

S

State

A representation of environment variables sufficient for a task or transition model, which may be measured, privileged, estimated, or latent.

T

Transition

A change from one state or observation to another, often represented as current context, action, next state, and outcome.

R

Rollout

A sequence generated by interacting with an environment or by repeatedly applying a model to imagine future states under actions.

T

Telemetry

Structured measurements emitted by a game, simulator, robot, or capture system, including state variables, events, timing, and system status.

T

Timebase

The clock and timestamp convention used to align streams within an episode or dataset.

D

Dataset provenance

The traceable history of where data came from, who collected or transformed it, which tools were used, and how a released file relates to its source.

O

Outcome

A result associated with a transition or episode, such as progress, reward, success, failure, event occurrence, or terminal condition.