Cognition · Data artifact

Reward Function Specification

Data artifactCognitionCognition & Memoryarc:RewardFunctionSpecification

A specification of the immediate payoff received on each state transition, together with the discount factor, from which long-term utility is derived.

Responsibility. Defines the reward signal that simulations produce and search optimizes.

Also known as: Terminal reward signal, Reward shaping, Reward function, Reward structure, Shaped reward, Potential function

Direct neighbourhood (hover for relationship types)

Relationships

configures structural

is configured by structural

Operational Norm Set Ch9.6

is read by dependency

is written by dependency

Inverse Reward Learner Ch5.10

is produced by lifecycle

Inverse Reward Learner Ch9.6

Design guidance

SHOULD directly reflect task objectives (e.g., +100 at goal, 0 otherwise) rather than intermediate progress.
If shaping is necessary, SHOULD reward progress only when strictly closer to the goal and cap total shaped reward.
SHOULD use potential-based shaping so dense guidance does not change which policy is optimal.
SHOULD include large penalties for actions the safety rules would override, so the learned policy internalizes constraints.

Quantitative guidance

As stated by the sources; verify before use.

Grid world reward: +100 goal, -100 trap, -1 per move (Ch5.12).

Classification

Patterns: Binary rewardsScalar rewardsVector (multi-objective) rewards with scalarizationWeighted-sum scalarizationPareto optimizationDiscounted rewardsReward engineeringPotential-based reward shaping (gamma*Phi(s') - Phi(s))Sparse vs. dense rewardConstraint-violation penalties
Risks mitigated: Reward misalignment / perverse incentives from shapingReward hackingSparse-reward learning failureNaive shaping changing the optimal policy

Sources

Ch5.5: T. Nguyen, "Monte Carlo Tree Search Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.5. ISBN: 9798244538229.
Ch5.10: T. Nguyen, "Utility-Based Decision Making Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.10. ISBN: 9798244538229.
Ch5.12: T. Nguyen, "Learning-Based Decision Making Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.12. ISBN: 9798244538229.
Ch5.13: T. Nguyen, "Hybrid Decision Systems Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.13. ISBN: 9798244538229.