PathNavigate: A Training-Free Pathology Agent with Surprise-Guided Scan and Shared Slide Memory for Whole-Slide Image VQA
arXiv:2605.23559v1 Announce Type: cross Abstract: Whole-slide image visual question answering (WSI-VQA) frames pathology as an extreme-context search problem: to answer a free-form clinical query, a system must first navigate a gigapixel slide under a strict inspection budget to locate sparse, high-resolution evidence. Existing approaches largely fall into two paradigms: i) supervised pathology multimodal large language models (MLLMs) and agents can absorb localization and reasoning into learned modules, but they often couple navigation to task-specific supervision and retraining, limiting the
The proliferation of gigapixel medical images and the increasing sophistication of AI models are driving demand for autonomous systems that can efficiently process and interpret complex visual data for diagnostic purposes.
This development addresses a critical bottleneck in pathology by offering an AI agent capable of navigating and interpreting whole-slide images with minimal human intervention, potentially accelerating diagnosis and improving accuracy.
Pathology analysis could shift towards more autonomous AI agents for initial screening and evidence discovery, reducing the burden on human pathologists and making expert analysis more scalable.
- · AI in healthcare sector
- · Pathology labs
- · Medical AI companies
- · Patients needing diagnostic services
- · Traditional pathology software vendors
- · Companies relying on manual image review in pathology
Pathologists will transition from primary image review to validating AI-generated findings and focusing on complex cases.
The cost and speed of diagnostic pathology could significantly improve, democratizing access to high-quality analysis.
This could set a precedent for AI agents in other extreme-context search problems across various scientific and industrial fields.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI