Model Adaptation · Software component
Reward Scorer
Software componentModel AdaptationModelsVariation point (abstract)arc:RewardScorer
An abstract component that computes the scalar reward signal for a prompt-response pair used to guide reinforcement-learning policy optimization.
Responsibility. Supplies reward signals to policy optimization.
Also known as: Reward function, Reward signal source
Variants
| Variant | When to choose |
|---|---|
| Composite Reward Scorer | Choose when human preference must be grounded by verifiable correctness, e.g., to prevent optimizing toward fluent hallucinations or annotator bias. |
| Constitutional Reward Scorer | Choose when harmlessness rewards must scale without human labels and reflect explicit, inspectable principles rather than implicit annotator preferences. |
| Ensemble Reward Scorer | Choose when reward hacking via single-pattern exploitation is a concern and the cost of training multiple reward models is acceptable. |
| Fairness-Constrained Reward Scorer | Choose for high-stakes decisions such as lending where the policy must satisfy a chosen fairness metric (e.g., equalized odds) while keeping legitimate business criteria. |
| Preference Reward Scorer | Choose when desired behaviour is a subjective judgment (e.g., tone, trade-offs) that rigid rules cannot capture and annotator bias is controlled. |
| Segment-Routed Reward Scorer | Choose when user segments legitimately hold divergent preferences and sufficient preference data can be collected for each segment. |
Relationships
hosts structural
is invoked by dependency
Design guidance
- SHOULD combine multiple principles (e.g., honesty alongside helpfulness) so one-sided optimizations are caught.
- SHOULD use a frozen reward model during RL optimization.
Classification
- Patterns
- Hybrid reward functions
- Quality attributes
- Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)
- Risks mitigated
- Reward hackingPrinciple gaming (safety and helpfulness principle hacking)
Sources
- Ch3.5: T. Nguyen, "Prompt Optimization, Few-Shot Learning, Fine-Tuning," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.5. ISBN: 9798244538229.
- Ch9.5: T. Nguyen, "Constitutional AI," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 9.5. ISBN: 9798244538229.
- Ch10.3: T. Nguyen, "RLHF Methodology," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 10.3. ISBN: 9798244538229.