mlsanityData to 7 Oct 2026

When Derived Measurements Mislead: Quantifying and Mitigating LLM Over-Trust with Privileged-Modality Reliability Evidence

arXiv 2607.28421 ↗First posted 30 Jul 2026cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Derived measurements increasingly enter large language model (LLM) pipelines as direct facts despite their instance-dependent validity. We define derived-feature over-trust (DFOT) as the failure in which a downstream LLM assigns such a measurement the epistemic status of a direct fact or uses it outside its valid scope. Using physiological sensing as a case study, D1 tests acceptance of a PPG-derived rhythm contradicted by offline ECG, whereas D2 tests rejection of an offline-confirmed reliable PPG rhythm under misleading severe history. ECG supplies training supervision and offline reference construction but is never shown to the LLM. Five estimands quantify this chain: conflict over-trust rate (COTR) and context-induced error rate (CIR) characterize D1/D2; correct repair rate (CRR) measures frozen-error repair; evidence-specific repair margin (ESRM) contrasts matched and patient-disjoint shuffled evidence; and utility harm rate (UHR) measures unnecessary verification among HIGH-reliability cases used without verification at baseline. The framework does not depend on a particular reliability generator. We demonstrate it on 50,000 paired PPG-ECG records using ECG-to-PPG privileged distillation as an illustrative baseline and PPG-only inference. On a protocol-locked 187-patient test, the baseline improves four repair and specificity endpoints by 1.82-6.69 percentage points, with all paired confidence intervals excluding zero; UHR increases by 0.67 percentage points (95% CI: -0.4 to +1.7). DFOT provides a common evaluation target for stronger mitigation methods. The code is available at https://github.com/Zongheng-Guo/When-Derived-Measurements-Mislead.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 32 of its 57 references that are in the map. First posted on arXiv 30 Jul 2026.

Builds on

32 of 57 references.

References · 32

  1. 2015Distilling the Knowledge in a Neural Networkcited by 10,076
  2. 2017On Calibration of Modern Neural Networkscited by 4,860
  3. 2016Simple and Scalable Predictive Uncertainty Estimation using Deep Ensemblescited by 3,888
  4. 2022Large Language Models Encode Clinical Knowledgecited by 1,807
  5. 2022Language Models (Mostly) Know What They Knowcited by 1,783
  6. 2019Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shiftcited by 1,274
  7. 2019Attention is not Explanationcited by 916
  8. 2017Selective Classification for Deep Neural Networkscited by 799
  9. 2023Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generationcited by 795
  10. 2023Visual ChatGPT: Talking, Drawing and Editing with Visual Foundation Modelscited by 659
  11. 2019Attention is not not Explanationcited by 629
  12. 2021Entity-Based Knowledge Conflicts in Question Answeringcited by 349
20 more
  1. 2019SelectiveNet: A Deep Neural Network with an Integrated Reject Optioncited by 333
  2. 2023AudioGPT: Understanding and Generating Speech, Music, Sound, and Talking Headcited by 264
  3. 2020Consistent Estimators for Learning to Defer to an Expertcited by 262
  4. 2017Predict Responsibly: Improving Fairness and Accuracy by Learning to Defercited by 228
  5. 2015Unifying distillation and privileged informationcited by 218
  6. 2015Cross Modal Distillation for Supervision Transfercited by 173
  7. 2022Rich Knowledge Sources Bring Complex Knowledge Conflicts: Recalibrating Models to Reflect Conflicting Evidencecited by 142
  8. 2024ClashEval: Quantifying the tug-of-war between an LLM's internal prior and external evidencecited by 109
  9. 2024PaPaGei: Open Foundation Models for Optical Physiological Signalscited by 64
  10. 2025OpenTSLM: Time-Series Language Models for Reasoning over Multivariate Medical Text- and Time-Series Datacited by 30
  11. 2022Toward Understanding Privileged Features Distillation in Learning-to-Rankcited by 16
  12. 2024PhysioLLM: Supporting Personalized Health Insights with Wearables and Large Language Modelscited by 16
  13. 2025Learning Game-Playing Agents with Generative Code Optimizationcited by 8
  14. 2026Understanding the Challenges in Iterative Generative Optimization with LLMscited by 8
  15. 2026SIGMA-PPG: Statistical-prior Informed Generative Masking Architecture for PPG Foundation Modelcited by 7
  16. 2025PPG-Distill: Efficient Photoplethysmography Signals Analysis via Foundation Model Distillationcited by 6
  17. 2026SP-Mind: An Autonomous Reasoning Agent for Spatial Proteomics Analysiscited by 4
  18. 2026A robust PPG foundation model using multimodal physiological supervisioncited by 3
  19. 2024ECG-guided individual identification via PPGcited by 1
  20. 2025BioX-Bridge: Model Bridging for Unsupervised Cross-Modal Knowledge Transfer across Biosignalscited by 1

Built on by

Not cited yet.