![[2602.18773] LAMMI-Pathology: A Tool-Centric Bottom-Up LVLM-Agent Framework for Molecularly Informed Medical Intelligence in Pathology](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.18773] LAMMI-Pathology: A Tool-Centric Bottom-Up LVLM-Agent Framework for Molecularly Informed Medical Intelligence in Pathology
Summary
The LAMMI-Pathology framework proposes a novel tool-centric approach for enhancing molecularly informed medical intelligence in pathology through scalable agent systems.
Why It Matters
This research addresses the limitations of traditional text-image diagnostic methods in pathology by introducing a more evidence-driven, tool-centric framework. As molecular techniques become more prevalent, LAMMI-Pathology could significantly improve diagnostic accuracy and efficiency in medical settings.
Key Takeaways
- LAMMI-Pathology utilizes a bottom-up architecture for agent tool-calling.
- The framework enhances inference robustness through a trajectory-aware fine-tuning strategy.
- Customized domain-adaptive tools are central to the framework's design.
- Atomic Execution Nodes (AENs) facilitate reliable agent-tool interactions.
- The approach aims to improve the accuracy of molecularly validated pathological diagnoses.
Computer Science > Artificial Intelligence arXiv:2602.18773 (cs) [Submitted on 21 Feb 2026] Title:LAMMI-Pathology: A Tool-Centric Bottom-Up LVLM-Agent Framework for Molecularly Informed Medical Intelligence in Pathology Authors:Haoyang Su, Shaoting Zhang, Xiaosong Wang View a PDF of the paper titled LAMMI-Pathology: A Tool-Centric Bottom-Up LVLM-Agent Framework for Molecularly Informed Medical Intelligence in Pathology, by Haoyang Su and 2 other authors View PDF HTML (experimental) Abstract:The emergence of tool-calling-based agent systems introduces a more evidence-driven paradigm for pathology image analysis in contrast to the coarse-grained text-image diagnostic approaches. With the recent large-scale experimental adoption of spatial transcriptomics technologies, molecularly validated pathological diagnosis is becoming increasingly open and accessible. In this work, we propose LAMMI-Pathology (LVLM-Agent System for Molecularly Informed Medical Intelligence in Pathology), a scalable agent framework for domain-specific agent tool-calling. LAMMI-Pathology adopts a tool-centric, bottom-up architecture in which customized domain-adaptive tools serve as the foundation. These tools are clustered by domain style to form component agents, which are then coordinated through a top-level planner hierarchically, avoiding excessively long context lengths that could induce task drift. Based on that, we introduce a novel trajectory construction mechanism based on Atomic Execution Nodes...
