Cognition · Software component
Reasoning Consistency Checker
Software componentCognitionCognition & Memoryarc:ContradictionDetector
A verification component that uses entailment judgments and entity resolution to detect contradictory assertions across steps of a reasoning chain.
Responsibility. Detects self-contradictions across reasoning steps.
Also known as: Coherence checker, Consistency checking across representations, Inter-step consistency checker, Contextual verification, Semantic consistency checking, Internal contradiction detection, Within-session consistency check
Relationships
is configured by structural
invokes dependency
is invoked by dependency
reads dependency
receives data from dynamic
sends data to dynamic
- Reasoning Quality Scorer abstract Ch3.9
guards control
- Answer Synthesizer abstract Ch3.10
evaluates assurance
is evaluated by assurance
Design guidance
- SHOULD use semantic entailment rather than keyword matching to recognise contradictions across different surface forms.
- SHOULD verify final outputs against all prior outputs in a multi-agent chain before customer communication.
- SHOULD combine relevance filtering, contradiction detection and source-quality assessment instead of naively inserting top-k results.
Quantitative guidance
As stated by the sources; verify before use.
- Contradiction threshold typically 0.6-0.7 (Ch3.9).
- Financial agent: 23% of reasoning chains contained contradictory statements before improvement (Ch3.9 case study).
- Consistency scores above 85% indicate reduced hallucination risk; below 70% flags instability (Ch3.10).
Classification
- Patterns
- Entailment-based contradiction detectionEntity resolutionInter-step consistency = 1 - max P_contradiction(prior information, new RCU)Inter-step consistency = 1 - contradictions / total stepsFlag conflicting claims across retrieved passages before context insertionConsistency verification of summaries against sources
- Quality attributes
- Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)Reliability (ISO/IEC 25010 | NIST AI RMF: valid and reliable)
- Risks mitigated
- Self-contradictionEntity conflationSelf-contradictory reasoning chainsLosing track of earlier conclusions in long chainsInconsistent answers across turnsCascading multi-agent hallucination
Sources
- Ch3.6: T. Nguyen, "Trace Analysis and Execution Debugging," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.6. ISBN: 9798244538229.
- Ch3.9: T. Nguyen, "Reasoning Quality," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.9. ISBN: 9798244538229.
- Ch3.10: T. Nguyen, "Efficiency Metrics," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.10. ISBN: 9798244538229.
- Ch5.9: T. Nguyen, "Working Memory," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.9. ISBN: 9798244538229.