MedHal-Loc: Are "Explainable-by-Architecture" Medical Hallucination Detectors Faithful Localizers? A Localization Benchmark

• URL: http://arxiv.org/abs/2606.21517v1
• Date: Fri, 19 Jun 2026 15:11:37 GMT
• Title: MedHal-Loc: Are "Explainable-by-Architecture" Medical Hallucination Detectors Faithful Localizers? A Localization Benchmark
• Abstract summary: We introduce MedHal-Loc, a benchmark and metric for localization faithfulness.<n>We find that NLI-per-clause, consistency-per-sentence, and the dedicated span detector FAVA all localize well above chance.
• Score: 3.738942365298833
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Detecting hallucinations in clinical text is increasingly framed as an explainability problem: systems should not merely flag an unreliable response but point to the offending span. Architectures built around knowledge-graph (KG) triple decomposition are marketed for exactly this auditability, yet their localization ability is typically assumed rather than measured. We introduce MedHal-Loc, a benchmark and metric for localization faithfulness -- whether a detector's top-ranked error unit actually overlaps the erroneous span. The controlled subset comprises 300 PubMedQA-derived statements with single, span-level errors injected across four localizable types (entity substitution, relation error, mechanism misattribution, invention), yielding gold spans by construction; a complementary natural subset documents that real hallucinations are dominated by diffuse conclusion-flips that resist span localization (a human expert accepted 1/18 candidate spans). Evaluating four fine-grained paradigms, we find that NLI-per-clause, consistency-per-sentence, and the dedicated span detector FAVA all localize well above chance, whereas an elaborate KG-triple pipeline localizes no better than chance (+3.3pp, n.s.), bottlenecked by ~59% entity-extraction coverage -- despite competitive detection F1 (0.609). Detection competence does not imply faithful localization; architectural explainability must be validated, not presumed.

Related papers

• MedSNIP: Building and Benchmarking Snippet-Level Granularity for Medical Fact Verification [50.13873960887466]
  We introduce MedSNIP-Bench, a benchmark for snippet-level medical fact verification.<n>We also introduce MedSNIP, an automatic snippet-generation pipeline.<n>Across MedSNIP-Bench, HealthFC, and MedHallu, snippet-level verification preserves or improves false-class F1.
  arXiv  Detail & Related papers  (2026-09-11T14:08:44Z)
• Enoki: Efficient Multi-Level Hallucination Detection [48.671558015589575]
  We propose Enoki, an Open Information Extraction framework for hallucination detection.<n>Enoki extracts text-anchored relational facts, verifies them against evidence, and projects unsupported facts back to hallucinated spans.<n>We show that Enoki remains competitive with strong claim-level systems while using fewer resources and achieves superior performance on fine-grained span- and entity-level localization.
  arXiv  Detail & Related papers  (2026-09-01T02:21:28Z)
• Tracing the Cascade: A Topology-Aware Evaluation Framework for Scientific Agent Hallucinations [27.330265411994258]
  SCHEMA is an evidence-grounded, topology-aware evaluation framework for hallucinations in scientific agents.<n>We show that hallucinations concentrate at a small set of highly connected knowledge hubs.<n>Results indicate that for high-stakes scientific applications, terminal accuracy alone is an insufficient signal of agent reliability.
  arXiv  Detail & Related papers  (2026-08-01T15:19:10Z)
• Inference-Time Steering for Cross-Lingual Factual Consistency in LLMs [51.56484100374058]
  Large Language Models (LLMs) demonstrate remarkable multilingual fluency, but their internal knowledge representations remain disproportionately biased toward high-resource languages.<n>We investigate whether these biases can be mitigated at inference time, forcing an English-prompted model to answer as if it were queried in target languages.<n>Experiments on Gemma 3 12B Instruct reveal persona prompting to be the strongest overall intervention, balancing efficacy, safety, and out-of-domain generalization.
  arXiv  Detail & Related papers  (2026-07-21T16:15:05Z)
• Per-Entity Bias Mapping for AI Visibility: Why Brand Mentions Require Entity-Specific Calibration [0.0]
  This paper argues that aggregate metrics are insufficient because entities exhibit systematically different AI visibility error profiles.<n>Three failure modes are identified: underrepresented entities suffer invisibility due to weak knowledge graph presence; large entities suffer the Brand Hallucination Paradox.
  arXiv  Detail & Related papers  (2026-06-19T16:50:29Z)
• CalBrief: A Pilot Diagnostic Benchmark for Evidence-Calibrated Scientific Briefing with Large Language Models [8.057966659934769]
  Large language models (LLMs) are increasingly used as research assistants, yet it remains unclear whether they can calibrate research takeaways to the strength and scope of the supporting evidence.<n>We study evidence-calibrated scientific briefing: given a bounded package of related papers, a system should generate package-level takeaways with evidence strength, scope boundaries, and missing-evidence caveats.<n>We use CalBrief, an auditable role/gap/strength framework, as a diagnostic probe to locate where briefing breaks down.
  arXiv  Detail & Related papers  (2026-06-11T06:49:52Z)
• R-APS: Compositional Reasoning and In-Context Meta-Learning for Constrained Design via Reflective Adversarial Pareto Search [18.8323743697237]
  Large language models (LLMs) are fluent on open-ended tasks, yet in agentic settings, fluency does not ensure reliable delivery.<n>We trace this gap to three coupled structural failures: errors propagate without localization, worst-case perturbations go unevaluated, and accumulated knowledge is never invalidated.<n>We argue these share a root cause: abductive, counterfactual, meta-inductive, corrective, and inductive reasoning pull a shared context in incompatible directions.
  arXiv  Detail & Related papers  (2026-06-03T12:45:39Z)
• When Should an AI Scientist Stop? Verifiable Experiment Steering and Refusal for Autonomous Discovery [0.0]
  CARTOGRAPH is a verification layer for AI scientists.<n>It couples unresolved-subspace experiment steering (select), explicit ambiguity closure (resolve), and residual-based library inadequacy detection (refuse)
  arXiv  Detail & Related papers  (2026-05-26T18:19:16Z)
• What Are We Actually Decoding? Source Attribution for Non-Invasive Brain-to-Language Retrieval [42.66754319854329]
  We recast stimulus-locked MEG-to-audio retrieval as an auditing framework.<n>We use structural shortcuts, window-level stimulus-locked evidence, and cross-window contextual aggregation.<n>These results suggest that brain-to-language performance should be source-attributed, not merely reported.
  arXiv  Detail & Related papers  (2026-05-23T11:23:39Z)
• Regulating Anatomy-Aware Rewards via Trajectory-Integral Feedback for Volumetric Computed Tomography Analysis [30.439440028190944]
  We introduce textbfClinical Abnormality Benchmarking Substrate (CABS), a structured system that decomposes radiology reports into verifiable clinical semantic units.<n>We propose textbfTrajectory-Integral FeedbackO (TIF-GRPO), a novel framework integrating control-theoretic principles into policy optimization.
  arXiv  Detail & Related papers  (2026-05-19T04:33:27Z)
• Discovery of Hidden Miscalibration Regimes [52.452902154360565]
  Models may be systematically overconfident on some kinds of inputs and underconfident on others.<n>We define the corresponding miscalibration field and propose a diagnostic framework for estimating it.<n>Our approach learns a calibration-aware representation of the input space and estimates signed local miscalibration by kernel smoothing in the learned geometry.
  arXiv  Detail & Related papers  (2026-05-13T13:07:50Z)
• Neural Uncertainty Principle: A Unified View of Adversarial Fragility and LLM Hallucination [60.197429875410286]
  Adversarial vulnerability in vision and hallucination in large language models are conventionally viewed as separate problems.<n>Formalizing a Neural Uncertainty Principle (NUP) under a loss-induced state, we find that in near-bound regimes, further compression must be accompanied by increased sensitivity dispersion.<n>In vision, masking highly coupled components improves robustness without costly adversarial training.<n>In language, the same prefill-stage probe detects hallucination risk before generating any answer tokens.
  arXiv  Detail & Related papers  (2026-03-20T02:07:10Z)
• Anatomy of a Lie: A Multi-Stage Diagnostic Framework for Tracing Hallucinations in Vision-Language Models [62.932580559941414]
  Vision-Language Models (VLMs) frequently "hallucinate" - generate plausible yet factually incorrect statements.<n>We propose a new paradigm for diagnosing hallucinations, recasting them from static output errors into dynamic pathologies of a model's computational cognition.
  arXiv  Detail & Related papers  (2026-03-16T17:20:38Z)
• The Semantic Illusion: Certified Limits of Embedding-Based Hallucination Detection in RAG Systems [0.0]
  We apply hallucination prediction to RAG detection, transforming scores into decision sets with finite-sample coverage guarantees.<n>We analyze this failure through the lens of distributional tails, showing that while NLI models achieve acceptable AUC (0.81), the "hardest" hallucinations are semantically indistinguishable from faithful responses.
  arXiv  Detail & Related papers  (2025-12-17T04:22:28Z)
• Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection [85.53263670166304]
  One-stage detector basically formulates object detection as dense classification and localization. Recent trend for one-stage detectors is to introduce an individual prediction branch to estimate the quality of localization. This paper delves into the representations of the above three fundamental elements: quality estimation, classification and localization.
  arXiv  Detail & Related papers  (2020-06-08T07:24:33Z)

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.