Difficulty: Advanced

Predicting protein 3D structure from sequence.

Variable whose values depend on chance.

Number of linearly independent rows or columns.

Interleaving reasoning and tool use.

Differences between simulated and real physics.

Simple agent responding directly to inputs.

Maximizing reward without fulfilling real goal.

Modifying reward to accelerate learning.

Motion of solid objects under forces.

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

Maintaining alignment under new conditions.

Sampling-based motion planner.

Systems where failure causes physical harm.

Using limited human feedback to guide large models.

AI applied to scientific problems.

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

Devices measuring physical quantities (vision, lidar, force, IMU, etc.).

Closed loop linking sensing and acting.

Ensuring AI allows shutdown.

Performance drop when moving from simulation to reality.

Trend reversal when data is aggregated improperly.

Artificial environment for training/testing agents.

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

Incremental capability growth.

Robots made of flexible materials.

Reward only given upon task completion.

Model exploits poorly specified objectives.

Inferring the agent’s internal state from noisy sensor data.

Decisions dependent on others’ actions.

Formal model linking causal mechanisms and variables.