Cognition · Software component

Exemplar Selector

Software componentCognitionCognition & MemoryVariation point (abstract)arc:ExemplarSelector

An abstract component that chooses which demonstration examples from a demonstration pool are placed into a prompt, and in what order, for in-context learning.

Responsibility. Composes the demonstration block used for few-shot prompting.

Also known as: Demonstration selector, Few-shot example selector

Direct neighbourhood (hover for relationship types)

Variants

Variant	When to choose
Cluster-Diverse Exemplar Selector	Choose when demonstrations must cover diverse problem types and reasoning chains are to be generated automatically rather than written manually.
Coreset Exemplar Pre-selector	Choose when the demonstration pool is large or noisy and a compact core subset satisfying sufficiency and necessity is needed.
Similarity Exemplar Selector	Choose when accuracy gains justify per-query retrieval cost; computational cost becomes high on large demonstration pools.

Relationships

is invoked by dependency

Prompt Context Builder Ch3.5

reads dependency

Prompt Exemplar Set Ch3.5

Design guidance

SHOULD start with 2-3 examples and add more only when evaluation shows clear improvement.
SHOULD place the most representative example last and order examples progressively from simple to complex.
SHOULD NOT rely on random sampling from available data.
SHOULD NOT add examples for tasks the model already handles zero-shot.

Quantitative guidance

As stated by the sources; verify before use.

Typical few-shot uses 2-5 demonstrations; performance plateaus or declines after 5-8 (Ch3.5).

Classification

Patterns: Few-shot in-context learningRecency-aware orderingSimple-to-complex ordering
Quality attributes: Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)Cost efficiencyPerformance efficiency (ISO/IEC 25010)
Risks mitigated: High variance from random demonstration selectionToken waste from excessive examplesRecency bias

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.