Cognition · Software component
Self-Consistency Aggregator
Software componentCognitionCognition & MemoryVariation point (abstract)arc:SelfConsistencyAggregator
An abstract cognition component that combines final answers from multiple independently generated reasoning chains or agents into one consensus answer by voting, exposing the agreement distribution.
Responsibility. Selects a consensus answer from multiple independent reasoning chains.
Also known as: Consensus aggregator, Answer vote aggregator, Ensemble voter
Variants
| Variant | When to choose |
|---|---|
| Majority Vote Aggregator | Choose for easy problems or small sample counts (k=3-5) where vote distributions are already clear and quality-assessment overhead is not justified. |
| Quality-Weighted Vote Aggregator | Choose for medium-to-hard problems with larger sample counts (k≈10-40) or cost-constrained high-stakes deployments where extracting more signal from fewer samples justifies ~5% quality-scoring overhead. |
| Test-Based Vote Aggregator | Choose for code generation where implementations vary textually but test cases define correctness. |
Relationships
writes dependency
receives data from dynamic
sends data to dynamic
Design guidance
- SHOULD surface divergent conclusions as uncertainty (request information, apply tie-breaking, or involve humans) rather than presenting false certainty.
- SHOULD be applied only where a definable correct answer exists; voting is meaningless for open-ended generation or summarisation.
- SHOULD preserve all sampled chains with the voted answer when audit defensibility is required.
Quantitative guidance
As stated by the sources; verify before use.
- Error overlap across prompts and sampling conditions was 69% higher than under independence (DiVeRSE); hallucinated drug interactions showed 87% error correlation across samples (Ch5.3).
- Medical coding with k=10 majority voting: 89.2% accuracy vs 71% CoT and 58% human coders; 94.3% inter-coding consistency; 10.8% of cases routed to human review (Ch5.3).
Classification
- Patterns
- Self-ConsistencyMajority votingMulti-agent consensusSelf-Consistency over Tree-of-Thoughts leavesSelf-Consistency at Graph-of-Thoughts refinement checkpointsSplit-vote detection
- Quality attributes
- Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)Reliability (ISO/IEC 25010 | NIST AI RMF: valid and reliable)Transparency and accountability (NIST AI RMF: accountable and transparent)
- Risks mitigated
- Path-specific reasoning errorsFalse certainty from a single reasoning chainOverconfident wrong answers
Sources
- Ch5.1: T. Nguyen, "Chain-of-Thought (CoT) Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.1. ISBN: 9798244538229.
- Ch5.3: T. Nguyen, "Self-Consistency Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.3. ISBN: 9798244538229.