Results for "Shapley values"

Measure of consistency across labelers; low agreement indicates ambiguous tasks or poor guidelines.

A formal privacy framework ensuring outputs do not reveal much about any single individual’s data contribution.

Standardized documentation describing intended use, performance, limitations, data, and ethical considerations.

Measures a model’s ability to fit random noise; used to bound generalization error.

Systematic error introduced by simplifying assumptions in a learning algorithm.

Allows model to attend to information from different subspaces simultaneously.

Estimating parameters by maximizing likelihood of observed data.

Stores past attention states to speed up autoregressive decoding.

Set of all actions available to the agent.

Fundamental recursive relationship defining optimal value functions.

Expected cumulative reward from a state or state-action pair.

Learning only from current policy’s data.

Learning from data generated by a different policy.

Sequential data indexed by time.

Persistent directional movement over time.

Decomposes a matrix into orthogonal components; used in embeddings and compression.

Measure of vector magnitude; used in regularization and optimization.

Describes likelihoods of random variable outcomes.

Average value under a distribution.

Measure of spread around the mean.

Visualization of optimization landscape.

Alternative formulation providing bounds.

Tendency for agents to pursue resources regardless of final goal.

Ensuring learned behavior matches intended objective.

Model behaves well during training but not deployment.

Learned subsystem that optimizes its own objective.

Maintaining alignment under new conditions.

Using limited human feedback to guide large models.

Review process before deployment.

Humans assist or override autonomous behavior.