Agentic data science with the Data Agent Kit
cloud.google.com·7h ago
AI Research
arXiv CS.AI·Sep 2·5 min read
SlideBank: A Persistent Hierarchical Evidence Bank for Consistent Whole-Slide Reasoning
TL;DR
SlideBank is a novel framework that organizes whole-slide images (WSIs) into a persistent, concept-indexed evidence bank, enhancing reasoning in pathology. It achieves over 52% accuracy in visual question answering tasks while significantly reducing inference costs through evidence reuse.
✦ Why It Matters
Implement SlideBank to enhance your pathology analysis and improve accuracy in visual question answering tasks.
Key Takeaways
- Utilize SlideBank for efficient whole-slide image analysis.
- Leverage hierarchical evidence banks for improved reasoning.
- Achieve high accuracy in visual question answering tasks.
- Reduce inference costs through persistent evidence reuse.
How It Works
SlideBank organizes whole-slide images into a persistent evidence bank, indexing concepts and grounding them spatially. It uses a coarse-to-fine exploration method to identify informative regions and converts these into explicit morphological observations.
During inference, it routes questions to relevant evidence and integrates information across different scales using a confidence-based consensus approach.