Agentic AI Reference Architecture

Model Adaptation · Software component

Inverse Reward Learner

Software componentModel AdaptationModelsarc:InverseRewardLearner

A preference elicitor that recovers the reward or utility function under which observed expert state-action behaviour would be optimal.

Responsibility. Infers a reward function from expert demonstrations via inverse reinforcement learning.

Also known as: Inverse reinforcement learning (IRL)

Variant of Preference Elicitor abstract

When to choose. Choose when expert demonstrations (e.g., thousands of hours of human driving) are available and explicit preference specification is impractical.

writes; producessends data toalternative toreadsspecializesReward Function Specification: writes; producesReward Function Specific…Domain Expert Annotator: sends data toDomain Expert AnnotatorRevealed Preference Learner: alternative toRevealed Preference Lear…Expert Demonstration Dataset: readsExpert Demonstration Dat…Preference Elicitor: specializesPreference Elicitor
Direct neighbourhood (hover for relationship types)

Relationships

reads dependency

writes dependency

sends data to dynamic

produces lifecycle

alternative to variability

Design guidance

Classification

Patterns
Inverse reinforcement learningPriors over plausible utility functionsActive learning queries to disambiguate hypothesesInverse Reinforcement LearningBottom-up value alignment

Sources

  1. 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.
  2. Ch9.6: T. Nguyen, "Value Alignment Frameworks," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 9.6. ISBN: 9798244538229.