paper-with-me

홈 › Papers

Shattering the Shortcut: A Topology-Regularized Benchmark for Multi-hop Medical Reasoning in LLMs

2026-03-12 · Xing Zi, Xinying Zhou, Jinghao Xiao, Catarina Moreira, Mukesh Prasad arxiv

While Large Language Models (LLMs) achieve expert-level performance on standard medical benchmarks through single-hop factual recall, they severely struggle with the complex, multi-hop diagnostic reasoning required in real-world clinical settings. A primary obstacle is "shortcut learning", where models exploit highly connected, generic hub nodes (e.g., "inflammation") in knowledge graphs to bypass authentic micro-pathological cascades. To address this, we introduce ShatterMed-QA, a bilingual benchmark of 10,558 multi-hop clinical questions designed to rigorously evaluate deep diagnostic reasoning. Our framework constructs a topology-regularized medical Knowledge Graph using a novel $k$-Shattering algorithm, which physically prunes generic hubs to explicitly sever logical shortcuts. We synthesize the evaluation vignettes by applying implicit bridge entity masking and topology-driven hard negative sampling, forcing models to navigate biologically plausible distractors without relying on superficial elimination. Comprehensive evaluations of 21 LLMs reveal massive performance degradation on our multi-hop tasks, particularly among domain-specific models. Crucially, restoring the masked evidence via Retrieval-Augmented Generation (RAG) triggers near-universal performance recovery, validating ShatterMed-QA's structural fidelity and proving its efficacy in diagnosing the fundamental reasoning deficits of current medical AI. Explore the dataset, interactive examples, and full leaderboards at our project website: https://shattermed-qa-web.vercel.app/

📄 PDF Abstract BibTeX arXiv:2603.12458

Code (0)

등록된 구현이 없습니다.

Tasks

Knowledge Graphs

Similar Papers 제목 키워드 기반

On a Sparse Shortcut Topology of Artificial Neural Networks

2018-11-22 · Fenglei Fan, Dayang Wang, Hengtao Guo, Qikui Zhu 외

In established network architectures, shortcut connections are often used to take the outputs of earlier layers as additional inputs to later layers. Despite the extraordinary effectiveness of shortcuts, there remain ope…

On the Complexity of Labeled Datasets

2019-11-13 · Rodrigo Fernandes de Mello

The Statistical Learning Theory (SLT) provides the foundation to ensure that a supervised algorithm generalizes the mapping $f: \mathcal{X} \to \mathcal{Y}$ given $f$ is selected from its search space bias $\mathcal{F}$.…

Learning Theory

An Improved Uniform Convergence Bound with Fat-Shattering Dimension

2023-07-13 · Roberto Colomboni, Emmanuel Esposito, Andrea Paudice

The fat-shattering dimension characterizes the uniform convergence property of real-valued functions. The state-of-the-art upper bounds feature a multiplicative squared logarithmic factor on the sample complexity, leavin…

The Cartesian Shortcut: Re-evaluate Vision Reasoning in Polar Coordinate Space

2026-05-11 · Xia Hu, Zhenrui Yue, Brian Potetz, Howard Zhou 외 arxiv

As current Multimodal Large Language Models rapidly saturate canonical visual reasoning benchmarks, a key question emerges: do these strong scores genuinely reflect robust visual understanding? We identify a pervasive vu…

Visual Reasoning

Norm-Hierarchy Transitions in Representation Learning: When and Why Neural Networks Abandon Shortcuts

2026-03-07 · Truong Xuan Khanh, Truong Quynh Hoa arxiv

Neural networks often rely on spurious shortcuts for many epochs before discovering structured representations. However, the mechanism governing when this transition occurs and whether its timing can be predicted remains…

Representation Learning