Technical primer

Action-conditioned data links decisions to futures

Understand the training primitive behind controllable prediction, interactive simulation, policy learning, and world action models.

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

Action-conditioned data records an observation history together with the control input that preceded a transition. It lets a model learn a distribution over possible next states given both context and action.

The action is not just another label

A control input changes which future should be predicted. The same scene can evolve differently if the agent accelerates, waits, turns, selects a tool, or issues a high-level command.

Representations must preserve semantics. Keyboard state, pointer delta, analog axis, discrete policy token, and robot motor command should not be flattened into one ambiguous field.

Alignment determines causality

If an action timestamp follows its visual consequence, the training sequence teaches the wrong order. Releases should declare the clock, capture latency, resampling method, execution delay, and uncertainty.

Commanded and executed actions can differ. Keep both when the environment can block, clip, delay, or override an input.

  • Raw input and normalized action
  • Issue time and effective time
  • Validity and missing-data masks
  • Environment and control-map version

Evaluate whether the model uses action

Reconstruction quality can look strong even when the model ignores control. Compare futures under alternative actions, measure event timing, and test long-horizon consistency.

Hold out action sequences and environments that expose memorization. Report uncertainty when several futures remain valid.

Questions this page should settle

What is an action trajectory?

It is a time-ordered record connecting observations, actions, state transitions, and outcomes across an episode.

How is action-conditioned data different from passive video?

Passive video shows what changed. Action-conditioned data also records the control input that preceded the change.

Do inferred actions count?

They can support research, but they should be labeled as model-generated estimates rather than silently treated as measured ground truth.

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