# Can LLMs Value the Right Evidence? Evidence-Value Misalignment in Dynamic Medical Diagnosis

arXiv: https://arxiv.org/abs/2609.35627 · HTML: https://mlsanity.com/paper/2609.35627/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 24 of its 35 references that are in the map.
- First posted on arXiv 28 Sep 2026.

## Builds on (most cited first, 24 of 24)

- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Language Models (Mostly) Know What They Know](https://mlsanity.com/paper/2207.05221/) (2022)
- [Large Language Models Encode Clinical Knowledge](https://mlsanity.com/paper/2212.13138/) (2022)
- [What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Exams](https://mlsanity.com/paper/2009.13081/) (2020)
- [PubMedQA: A Dataset for Biomedical Research Question Answering](https://mlsanity.com/paper/1909.06146/) (2019)
- [DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence](https://mlsanity.com/paper/2606.19348/) (2026)
- [Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs](https://mlsanity.com/paper/2306.13063/) (2023)
- [MedMCQA : A Large-scale Multi-Subject Multi-Choice Dataset for Medical domain Question Answering](https://mlsanity.com/paper/2203.14371/) (2022)
- [Kimi K2.5: Visual Agentic Intelligence](https://mlsanity.com/paper/2602.02276/) (2026)
- [LLMs Get Lost In Multi-Turn Conversation](https://mlsanity.com/paper/2505.06120/) (2025)
- [Robots That Ask For Help: Uncertainty Alignment for Large Language Model Planners](https://mlsanity.com/paper/2307.01928/) (2023)
- [Selective Question Answering under Domain Shift](https://mlsanity.com/paper/2006.09462/) (2020)
- [AgentClinic: a multimodal agent benchmark to evaluate AI in simulated clinical environments](https://mlsanity.com/paper/2405.07960/) (2024)
- [MediQ: Question-Asking LLMs and a Benchmark for Reliable Interactive Clinical Reasoning](https://mlsanity.com/paper/2406.00922/) (2024)
- [AI Hospital: Benchmarking Large Language Models in a Multi-agent Medical Interaction Simulator](https://mlsanity.com/paper/2402.09742/) (2024)
- [DDXPlus: A New Dataset For Automatic Medical Diagnosis](https://mlsanity.com/paper/2205.09148/) (2022)
- [DiagnosisArena: Benchmarking Diagnostic Reasoning for Large Language Models](https://mlsanity.com/paper/2505.14107/) (2025)
- [EHRWorld: A Patient-Centric Medical World Model for Long-Horizon Clinical Trajectories](https://mlsanity.com/paper/2602.03569/) (2026)
- [Knowing When to Abstain: Medical LLMs Under Clinical Uncertainty](https://mlsanity.com/paper/2601.12471/) (2026)
- [ChronoMedicalWorld: A Medical World Model for Learning Patient Trajectories from Longitudinal Care Data](https://mlsanity.com/paper/2605.21963/) (2026)
- [MedEinst: Benchmarking the Einstellung Effect in Medical LLMs through Counterfactual Differential Diagnosis](https://mlsanity.com/paper/2601.06636/) (2026)
- [Benchmarking Multi-turn Medical Diagnosis: Hold, Lure, and Self-Correction](https://mlsanity.com/paper/2604.04325/) (2026)
- [Quantifying and Mitigating Premature Closure in Frontier LLMs](https://mlsanity.com/paper/2605.15000/) (2026)
- [MedAction: Towards Active Multi-turn Clinical Diagnostic LLMs](https://mlsanity.com/paper/2605.07305/) (2026)

## Abstract (arXiv)

A correct diagnosis reached from insufficient or misleading evidence can pose a clinical hazard, yet outcome-based accuracy may reward such lucky guesses. We call this mismatch between diagnostic decisions and the value of available evidence Evidence-Value Misalignment (EVM). To disentangle evidential grounding independently from diagnostic accuracy, we introduce MedEVM, a dynamic benchmarking environment comprising 1,050 cases across 24 disease systems. Observations arrive turn by turn, requiring models to continuously calibrate its decision by deciding whether to wait for more evidence or submit a diagnosis. Across 9 LLMs, four interesting patterns are observed. (1) Miscalibrated evidence tracking. Making a diagnosis often fails to calibrate evidence sufficiency, even in more capable models, and even worsens in reasoning mode. (2) Misaligned diagnosis submission. Confidence in the correct diagnosis often fails to ensure timely submission despite sufficient evidence. (3) Evidence order matters. Reordering the same evidence changes diagnoses even when model confidence remains similar. (4) Misleading evidence remains influential. Added misleading evidence redirects diagnoses even after prior evidence becomes sufficient. We further verify that EVM predicts errors and that preventing premature submission improves accuracy. These findings motivate Evidence-Verified Diagnosis Harness (EVD-Harness). It decouples diagnosis generation from submission through an offline Contrastive Diagnostic Wiki and three online control stages, namely observation management, proposal and witness verification, and diagnosis submission control. Across five LLMs, EVD-Harness improves accuracy by 12.0--51.1 percentage points while mitigating EVM-related failures. Our results demonstrate that verifying evidential support before submission can make diagnostic decisions more reliable.