# When Confidence Takes the Wrong Path: Diagnosing Retrieval-State Lock-In in RAG

arXiv: https://arxiv.org/abs/2606.22728 · HTML: https://mlsanity.com/paper/2606.22728/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in RAG robustness and adaptive retrieval, one of 15 directions in Retrieval-augmented generation.
- Builds on 27 of its 82 references that are in the map.
- First posted on arXiv 22 Jun 2026.

## Placed in

- Retrieval-augmented generation / [RAG robustness and adaptive retrieval](https://mlsanity.com/rag/rag-robustness-and-adaptive-retrieval/)

## Builds on (most cited first, 27 of 27)

- [Language Models (Mostly) Know What They Know](https://mlsanity.com/paper/2207.05221/) (2022)
- [From Local to Global: A Graph RAG Approach to Query-Focused Summarization](https://mlsanity.com/paper/2404.16130/) (2024)
- [TrustLLM: Trustworthiness in Large Language Models](https://mlsanity.com/paper/2401.05561/) (2024)
- [Corrective Retrieval Augmented Generation](https://mlsanity.com/paper/2401.15884/) (2024)
- [ClashEval: Quantifying the tug-of-war between an LLM's internal prior and external evidence](https://mlsanity.com/paper/2404.10198/) (2024)
- [Uncertainty-Based Abstention in LLMs Improves Safety and Reduces Hallucinations](https://mlsanity.com/paper/2404.10960/) (2024)
- [RAG vs. GraphRAG: A Systematic Evaluation and Key Insights](https://mlsanity.com/paper/2502.11371/) (2025)
- [When to use Graphs in RAG: A Comprehensive Analysis for Graph Retrieval-Augmented Generation](https://mlsanity.com/paper/2506.05690/) (2025)
- [Towards Trustworthy Retrieval Augmented Generation for Large Language Models: A Survey](https://mlsanity.com/paper/2502.06872/) (2025)
- [How well do LLMs cite relevant medical references? An evaluation framework and analyses](https://mlsanity.com/paper/2402.02008/) (2024)
- [Semantic Volume: Quantifying and Detecting both External and Internal Uncertainty in LLMs](https://mlsanity.com/paper/2502.21239/) (2025)
- [UncertaintyRAG: Span-Level Uncertainty Enhanced Long-Context Modeling for Retrieval-Augmented Generation](https://mlsanity.com/paper/2410.02719/) (2024)
- [Fine-Grained Uncertainty Decomposition in Large Language Models: A Spectral Approach](https://mlsanity.com/paper/2509.22272/) (2025)
- [SURE-RAG: Sufficiency and Uncertainty-Aware Evidence Verification for Selective Retrieval-Augmented Generation](https://mlsanity.com/paper/2605.03534/) (2026)
- [SUGAR: Leveraging Contextual Confidence for Smarter Retrieval](https://mlsanity.com/paper/2501.04899/) (2025)
- [Semantic Uncertainty Quantification of Hallucinations in LLMs: A Quantum Tensor Network Based Method](https://mlsanity.com/paper/2601.20026/) (2026)
- [Faithfulness-Aware Uncertainty Quantification for Fact-Checking the Output of Retrieval Augmented Generation](https://mlsanity.com/paper/2505.21072/) (2025)
- [Uncertainty Quantification for Retrieval-Augmented Reasoning](https://mlsanity.com/paper/2510.11483/) (2025)
- [NOVA: NOise-aware Verbal Confidence CAlibration for Robust Large Language Models in RAG Systems](https://mlsanity.com/paper/2601.11004/) (2026)
- [Revisiting RAG Retrievers: An Information Theoretic Benchmark](https://mlsanity.com/paper/2602.21553/) (2026)
- [Augmenting Textual Generation via Topology Aware Retrieval](https://mlsanity.com/paper/2405.17602/) (2024)
- [Comparing RAG and GraphRAG for Page-Level Retrieval Question Answering on a Math Textbook](https://mlsanity.com/paper/2509.16780/) (2025)
- [When to Trust: A Causality-Aware Calibration Framework for Accurate Knowledge Graph Retrieval-Augmented Generation](https://mlsanity.com/paper/2601.09241/) (2026)
- [INTRYGUE: Induction-Aware Entropy Gating for Reliable RAG Uncertainty Estimation](https://mlsanity.com/paper/2603.21607/) (2026)
- [Role Support in Knowledge-Graph Error Ranking: Predictor Regimes and Evaluation Policies](https://mlsanity.com/paper/2512.22318/) (2025)
- [L-RAG: Balancing Context and Retrieval with Entropy-Based Lazy Loading](https://mlsanity.com/paper/2601.06551/) (2026)
- [Uncertainty-Aware Dynamic Knowledge Graphs for Reliable Question Answering](https://mlsanity.com/paper/2601.09720/) (2025)

## Built on by (most cited first, 2 of 2)

- [Uncertainty Quantification for LLM Agents: A Taxonomy, an Evaluation Protocol, and an Empirical Study](https://mlsanity.com/paper/2609.07395/) (2026)
- [Relevance Is Not Sufficient Evidence: Detecting Evidence Gaps Before Generation in RAG](https://mlsanity.com/paper/2609.37469/) (2026)

## Abstract (arXiv)

The trustworthiness of a retrieval-augmented generation (RAG) system depends on more than the answer it returns, yet many black-box uncertainty methods still read agreement among sampled answers as confidence. That inference fails when repeated samples condition on the same defective retrieval state. The state may be empty, with the model falling back on parametric memory, or populated by a coherent but wrong neighbourhood. In either case, the answers agree because the error is stable. The problem is recognised in deployed RAG, but it has lacked a name, a measurable signature, and a prevalence bound. We supply all three. We name the failure retrieval-state lock-in and diagnose it by separating the three objects a single confidence score conflates: the answer surface, the retrieved evidence, and the retrieval state itself. In an inspectable, ontology-guided knowledge-graph RAG (KG-RAG) system across six question-answering snapshots, we measure the agreement blind spot directly: at five samples per question, 42% of KG-RAG errors and 59% of dense-retrieval errors carry zero answer dispersion, so agreement has nothing to rank, while evidence- and retrieval-state checks still flag most of them. The decomposition supports an auditable decision rule: accepting an answer only when answer, evidence, and retrieval checks all agree that it is low-risk reaches 91.9% pooled precision against a 69.7% accept-all rate. The cost is coverage: it certifies only 7.7% of answers as low-risk. On the clinical calibration domain it reaches 100% precision under an automated judge; this is an in-domain automated-label upper bound, not a clinical safety claim, and still needs human validation. Confidence in RAG is object-specific: when answers agree, the useful question is which part of the pipeline to distrust.