# The Knowledge-Reasoning Dissociation: Fundamental Limitations of LLMs in Clinical Natural Language Inference

arXiv: https://arxiv.org/abs/2508.10777 · HTML: https://mlsanity.com/paper/2508.10777/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Clinical and biomedical applications, one of 17 directions in LLM reasoning.
- Builds on 25 of its 54 references that are in the map.
- First posted on arXiv 14 Aug 2025.

## Placed in

- LLM reasoning / [Clinical and biomedical applications](https://mlsanity.com/lr/clinical-and-biomedical-applications/)

## Builds on (most cited first, 25 of 25)

- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Scaling Laws for Neural Language Models](https://mlsanity.com/paper/2001.08361/) (2020)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [GPT-4o System Card](https://mlsanity.com/paper/2410.21276/) (2024)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [Training Compute-Optimal Large Language Models](https://mlsanity.com/paper/2203.15556/) (2022)
- [Sparks of Artificial General Intelligence: Early experiments with GPT-4](https://mlsanity.com/paper/2303.12712/) (2023)
- [Annotation Artifacts in Natural Language Inference Data](https://mlsanity.com/paper/1803.02324/) (2018)
- [CLUTRR: A Diagnostic Benchmark for Inductive Reasoning from Text](https://mlsanity.com/paper/1908.06177/) (2019)
- [Disentangled Representation Learning with Large Language Models for Text-Attributed Graphs](https://mlsanity.com/paper/2310.18152/) (2023)
- [LLMs for Generating and Evaluating Counterfactuals: A Comprehensive Study](https://mlsanity.com/paper/2405.00722/) (2024)
- [Logically Consistent Language Models via Neuro-Symbolic Integration](https://mlsanity.com/paper/2409.13724/) (2024)
- [Autoformalization in the Wild: Assessing LLMs on Real-World Mathematical Definitions](https://mlsanity.com/paper/2502.12065/) (2025)
- [Questioning Representational Optimism in Deep Learning: The Fractured Entangled Representation Hypothesis](https://mlsanity.com/paper/2505.11581/) (2025)
- [SemEval-2024 Task 2: Safe Biomedical Natural Language Inference for Clinical Trials](https://mlsanity.com/paper/2404.04963/) (2024)
- [Mitigating Content Effects on Reasoning in Language Models through Fine-Grained Activation Steering](https://mlsanity.com/paper/2505.12189/) (2025)
- [SemEval-2023 Task 7: Multi-Evidence Natural Language Inference for Clinical Trial Data](https://mlsanity.com/paper/2305.02993/) (2023)
- [Multi-Operational Mathematical Derivations in Latent Space](https://mlsanity.com/paper/2311.01230/) (2023)
- [Enhancing Large Language Models through Neuro-Symbolic Integration and Ontological Reasoning](https://mlsanity.com/paper/2504.07640/) (2025)
- [StructTest: Benchmarking LLMs' Reasoning through Compositional Structured Outputs](https://mlsanity.com/paper/2412.18011/) (2024)
- [Latent Concept Disentanglement in Transformer-based Language Models](https://mlsanity.com/paper/2506.16975/) (2025)
- [Discovering Highly Influential Shortcut Reasoning: An Automated Template-Free Approach](https://mlsanity.com/paper/2312.09718/) (2023)
- [ACR: A Benchmark for Automatic Cohort Retrieval](https://mlsanity.com/paper/2406.14780/) (2024)
- [Learning to Disentangle Latent Reasoning Rules with Language VAEs: A Systematic Study](https://mlsanity.com/paper/2506.19418/) (2025)
- [Reasoning in Neurosymbolic AI](https://mlsanity.com/paper/2505.20313/) (2025)

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

- [Compartmentalised Agentic Reasoning for Clinical NLI](https://mlsanity.com/paper/2509.10222/) (2025)
- [CLExEval: A Human-in-the-Loop Framework for Qualitative Evaluation of LLM Clinical Reasoning](https://mlsanity.com/paper/2606.31608/) (2026)

## Abstract (arXiv)

Large language models are often assumed to acquire increasingly structured, generalizable internal representations simply by scaling data and parameters. We interrogate this assumption by introducing a Clinical Trial Natural Language Inference benchmark comprising four reasoning families, Causal Attribution, Compositional Grounding, Epistemic Verification, and Risk State Abstraction. Each item is paired with a targeted Ground Knowledge and Meta-Level Reasoning Verification (GKMRV) probe, allowing us to dissociate failures of factual access from failures of inference. We evaluate six contemporary LLMs under both direct and chain of thought prompting. Models achieve near-ceiling GKMRV accuracy (mean accuracy 0.918) yet perform poorly on the main reasoning tasks (mean accuracy 0.25). Despite low accuracy, output inferences are highly consistent across samples (mean 0.87), indicating a systematic application of underlying heuristics and shortcuts. These results reveal fundamental structural and representational limitations: current LLMs often possess the relevant clinical knowledge but lack the structured, composable internal representations needed to deploy it reliably (e.g., integrating constraints, weighing evidence, or simulating counterfactuals). Decoupling knowledge from reasoning with GKMRV makes this dissociation explicit and measurable, providing an effective framework for probing the reliability of LLMs in high-stakes domains.