Results for "intuitive dynamics"

Human-like understanding of physical behavior.

Motion of solid objects under forces.

Motion considering forces and mass.

Predicts next state given current state and action.

Modeling environment evolution in latent space.

Modeling interactions with environment.

Collective behavior without central control.

Optimizes future actions using a model of dynamics.

Equations governing how system states change over time.

RL without explicit dynamics model.

Physical form contributes to computation.

The physical system being controlled.

Software simulating physical laws.

RL using learned or known environment models.

Learned model of environment dynamics.

Competition arises without explicit design.

Agents copy others’ actions.

A scalar measure optimized during training, typically expected loss over data, sometimes with regularization terms.

Methods to set starting weights to preserve signal/gradient scales across layers.

The text (and possibly other modalities) given to an LLM to condition its output behavior.

A high-priority instruction layer setting overarching behavior constraints for a chat model.

Set of all actions available to the agent.

Generative model that learns to reverse a gradual noise process.

Learns the score (∇ log p(x)) for generative sampling.

Pixel motion estimation between frames.

Temporal and pitch characteristics of speech.

Sequential data indexed by time.

Competitive advantage from proprietary models/data.

Declining differentiation among models.

Field combining mechanics, control, perception, and AI to build autonomous machines.