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

• URL: http://arxiv.org/abs/2606.22728v1
• Date: Mon, 22 Jun 2026 00:08:00 GMT
• Title: When Confidence Takes the Wrong Path: Diagnosing Retrieval-State Lock-In in RAG
• Abstract summary: Many black-box uncertainty methods still read agreement among sampled answers as confidence.<n>The problem is recognised in deployed RAG, but it has lacked a name, a measurable signature, and a prevalence bound.<n>We name the failure retrieval-state lock-in and diagnose it by separating the three objects a single confidence score conflates.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: 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.

Related papers

• Claim-Level Confidence Calibration for Reliable Decision Making with Large Language Models [7.714704554667688]
  Large Language Models (LLMs) increasingly support decision-making in high-stakes domains.<n>LLMs often hallucinate and express confidence that is misaligned with factual correctness.<n>We study claim-level confidence calibration as a decision-relevant uncertainty signal.
  arXiv  Detail & Related papers  (2026-08-23T16:09:54Z)
• Wrong but Useful: Trajectory Value Beyond Answer Correctness in Multi-Agent Messages [44.340269814027174]
  Multi-agent reasoning systems often use agreement, confidence, or automated scores to decide which messages should shape a final answer.<n>Yet a wrong answer can contain a useful decomposition, constraint, or scientific principle.<n>We test this distinction with Diverse Hypothesis Deliberation (DHD), a controlled measurement protocol that caches five independently generated messages.
  arXiv  Detail & Related papers  (2026-08-14T15:16:57Z)
• When Many Answers Are Valid, Voting Fails: Symbolic Verification for Best-of-K Causal Reasoning in LLMs [42.32694162421078]
  We introduce CALVER, a training-free symbolic verifier that scores structured traces against Pearl's causal criteria.<n>On CLEAR find-one-valid queries that admit multiple graph-valid answers, CALVER reaches 42.1% where plurality, a reward model, an LLM judge, and model confidence remain near 30%.
  arXiv  Detail & Related papers  (2026-08-04T11:45:46Z)
• Two Regimes of Chain-of-Thought Unfaithfulness: Behavioral Detection Fails Where Models Are Wrong [1.2744523252873352]
  Black-box (behavioral) detection of unfaithful CoT against FaithCoT-Bench's human annotations.<n>We find answer correctness structures the problem at every level.
  arXiv  Detail & Related papers  (2026-07-26T04:43:53Z)
• Catching One in Five: LLM-as-Judge Blind Spots in Production Multi-Turn Transaction Agents [45.148328075418156]
  We study a deployed multi-turn food-and-beverage ordering agent and measure how many genuine quality problems it catches.<n>Our blind-spot taxonomy shows the failure is structured, not random.<n>For production multi-turn agents, automated judging is a regression floor, not a substitute for human review.
  arXiv  Detail & Related papers  (2026-06-09T02:11:01Z)
• Risk-Controlled Lean-as-Judge for Natural-Language Mathematical Reasoning [16.398764816978584]
  Lean is increasingly used to judge natural-language mathematical answers, but its signal is partial.<n>We show this signal is (i) sharply coverage-dependent, that is the proof-winning answer is correct 96% of the time at high proved coverage but 20% at low.<n>A 7B autoformalizer proves a class for only 28% of problems, and a manual audit finds only approximately 43% of those faithful.<n>We propose COVCAL, a selector over Lean-trace diagnostics that certifies a finite-sample selective-risk bound on accepted answers or abstains.
  arXiv  Detail & Related papers  (2026-05-27T11:59:28Z)
• MAIGO: Mitigating Lost-in-Conversation with History-Cleaned On-Policy Self-Distillation [51.32912774464992]
  Large language models often solve tasks from a fully specified prompt but degrade when the same requirements unfold over multiple turns.<n>We propose MAIGO, an on-policy self-distillation method that reduces this contamination using history-cleaned references from the model's own policy.<n>For middle turns, MAIGO removes prior assistant replies while preserving the user-visible sharded prefix; for answer turns, it distills from paired full-view references conditioned on the completed user-side dialogue.
  arXiv  Detail & Related papers  (2026-05-26T15:38:46Z)
• Claim-Selective Certification for High-Risk Medical Retrieval-Augmented Generation [0.0]
  Mixed evidence may support one claim, require conditions for another, and contradict a third.<n>We study claimselective responses into verifiable claims, scored against retrieved evidence, and mapped by an intent-aware selector to full, partial, conflict, abstain<n>The resulting interface separates action-label prediction from evidence-linked claim selection under mixed evidence.
  arXiv  Detail & Related papers  (2026-05-21T03:29:50Z)
• Distributional Energy-Based Models for Uncertainty-Aware Structured LLM Reasoning [40.342912574072024]
  Large Language Models produce structured outputs such as travel plans or code solutions.<n>Individual reasoning steps may appear correct while the output as a whole violates budgets, fails test cases, or contradicts earlier deductions.<n>We propose a learned quality scorer with deterministic analytical constraint penalties for verifying structured LLM outputs.
  arXiv  Detail & Related papers  (2026-05-15T17:08:27Z)
• Pause and Reflect: Conformal Aggregation for Chain-of-Thought Reasoning [8.024041325202612]
  Chain-of-thought (CoT) reasoning with self-consistency improves performance by aggregating multiple sampled reasoning paths.<n>We introduce a conformal procedure for CoT reasoning that directly addresses aggregation uncertainty.<n>Our approach replaces majority voting with weighted score aggregation over reasoning paths and calibrates an abstention rule using conformal risk control.
  arXiv  Detail & Related papers  (2026-05-13T20:33:59Z)
• The Last Word Often Wins: A Format Confound in Chain-of-Thought Corruption Studies [0.0]
  Corruption studies infer which steps are computationally important'' from accuracy loss when steps are corrupted.<n>We show that when benchmark chains end with an explicit terminal answer line, these tests largely measure emphanswer placement rather than where intermediate computation is carried out.
  arXiv  Detail & Related papers  (2026-05-11T16:26:50Z)
• Trust or Abstain? A Self-Aware RAG Approach [45.1230482333942]
  Retrieval-augmented generation (RAG) improves large language models by incorporating external evidence.<n>Existing approaches mainly coordinate which source to use, without explicitly asking whether each answer path is correct.<n>We argue that faithful RAG requires LLM self-awareness, namely the ability to recognize the limits of its own knowledge and reasoning.
  arXiv  Detail & Related papers  (2026-05-11T05:44:27Z)
• SURE-RAG: Sufficiency and Uncertainty-Aware Evidence Verification for Selective Retrieval-Augmented Generation [6.604874054866016]
  We present SURE-RAG, a transparent aggregation protocol built on the observation that evidence sufficiency is a set-level property.<n>A shared pair-level claim-evidence verifier produces local relation distributions, which SURE-RAG aggregates into interpretable answer-level signals.<n>We evaluate on HotpotQA-RAG v3, a controlled multi-hop benchmark, under an artifact-aware protocol.
  arXiv  Detail & Related papers  (2026-05-05T09:05:40Z)
• Decomposed Prompting Does Not Fix Knowledge Gaps, But Helps Models Say "I Don't Know" [47.930782177987446]
  Large language models often struggle to recognize their knowledge limits in closed-book question answering, leading to confident hallucinations.<n>We evaluate three task-equivalent prompting regimes: Direct, Assistive, and Incremental, across different model scales and multi-hop QA benchmarks.<n>Because factual knowledge is stable while hallucinations are agreement, cross-regime provides a precise signal of internal uncertainty.
  arXiv  Detail & Related papers  (2026-02-04T18:39:58Z)
• CARE What Fails: Contrastive Anchored-REflection for Verifiable Multimodal [84.71254539482369]
  Group-relative reinforcement learning with verifiable rewards (RLVR) often wastes the most informative data it already has the failures.<n>We present CARE, a failure-centric post-training framework for multimodal reasoning that turns errors into supervision.<n> CARE improves accuracy and training smoothness while explicitly increasing the share of learning signal that comes from failures.
  arXiv  Detail & Related papers  (2025-12-22T16:34:21Z)
• Selective Question Answering under Domain Shift [90.021577320085]
  Abstention policies based solely on the model's softmax probabilities fare poorly, since models are overconfident on out-of-domain inputs. We train a calibrator to identify inputs on which the QA model errs, and abstain when it predicts an error is likely. Our method answers 56% of questions while maintaining 80% accuracy; in contrast, directly using the model's probabilities only answers 48% at 80% accuracy.
  arXiv  Detail & Related papers  (2020-06-16T19:13:21Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.