Agentic AI Reference Architecture

Knowledge & Data · Software component

Reranker

Software componentKnowledge & DataKnowledge & DataVariation point (abstract)arc:Reranker

A software component that deduplicates and re-scores candidate results from one or more retrieval sources into a single ranking.

Responsibility. Re-scores and orders candidate results into one ranking.

Also known as: Result Merger, Result Fusion, Reranking model, Cross-encoder reranker

receives data fromis orchestrated byreceives data fromis monitored bysends data toreceives data fromsends data tois configured byis invoked byreceives data fromis specialized bysends data tois specialized byis invoked byis specialized byhostsRetriever: receives data fromRetrieverRAG Query Orchestrator: is orchestrated byRAG Query OrchestratorVector Retriever: receives data fromVector RetrieverGraceful Degradation Manager: is monitored byGraceful Degradation Man…Context Assembler: sends data toContext AssemblerDense-Sparse Hybrid Retriever: receives data fromDense-Sparse Hybrid Retr…Context Compressor: sends data toContext CompressorKnowledge Chunk Metadata Schema: is configured byKnowledge Chunk Metadata…Parallel Fusion Retriever: is invoked byParallel Fusion RetrieverKeyword Retriever: receives data fromKeyword RetrieverMulti-Signal Relevance Ranker: is specialized byMulti-Signal Relevance R…Retrieval Confidence Filter: sends data toRetrieval Confidence Fil…Cross-Modal Reranker: is specialized byCross-Modal RerankerCoarse-to-Fine Vector Retriever: is invoked byCoarse-to-Fine Vector Re…Lexical Reranker: is specialized byLexical RerankerCross-Encoder Reranking Model: hostsCross-Encoder Reranking …
Direct neighbourhood (hover for relationship types)

Variants

VariantWhen to choose
Cross-Modal Reranker—
Lexical RerankerChoose instead of full hybrid fusion for extremely latency-sensitive applications (sub-50 ms p99) that still need keyword precision.
Multi-Signal Relevance Ranker—

Relationships

hosts structural

is configured by structural

is invoked by dependency

receives data from dynamic

sends data to dynamic

is orchestrated by control

is monitored by assurance

Design guidance

Quantitative guidance

As stated by the sources; verify before use.

Classification

Patterns
Score fusion (semantic score + graph proximity)Cross-encoder semantic re-rankingMulti-stage selective retrieval, stage 2: rerank by query intent, quality signals, recency and source authority to ~top-10Deduplication of redundant passagesFull-dimensional embedding rerank of shortlisted candidatesOn-the-fly GPU reranking of thousands of candidatesRecency-based rankingCross-encoder rerankingLLM listwise reranking
Technologies
LangChain ContextualCompressionRetrieverLlama 3.2 RerankQA 1B V2
Quality attributes
Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)

Sources

  1. Ch1.7B: T. Nguyen, "Relational Reasoning with Knowledge Graphs - Hybrid RAG+KG Integration," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 1.7B. ISBN: 9798244538229.
  2. Ch3.4: T. Nguyen, "Tuning Model Parameters for Production Performance," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.4. ISBN: 9798244538229.
  3. 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.
  4. 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.
  5. Ch6.1: T. Nguyen, "Embeddings and RAG Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 6.1. ISBN: 9798244538229.
  6. Ch6.4: T. Nguyen, "Data Quality Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 6.4. ISBN: 9798244538229.
  7. Ch6.5: T. Nguyen, "Production RAG Systems," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 6.5. ISBN: 9798244538229.
  8. Ref7.07: E. Li, V. Bellotti, R. Kraus, and R. Kao, "Build a retrieval-augmented generation (RAG) agent with NVIDIA Nemotron," NVIDIA Technical Blog, Sep. 23, 2025. [Online]. Available: https://developer.nvidia.com/blog/build-a-rag-agent-with-nvidia-nemotron/