LaGuadia: Language-Guided Adaptive Distillation from Pathology Foundation Models

2026-07-13

AI Summary

This work addresses the high computational cost of existing foundation models in computational pathology and the limitations of conventional knowledge distillation methods, which rely on static, uniform weighting and overlook tissue heterogeneity and clinical semantics. To overcome these issues, the authors propose a clinical language–guided adaptive multi-teacher knowledge distillation framework that introduces pathological report keywords as semantic anchors for the first time. The framework leverages a vision–language meta-teacher (MedSigLIP) to achieve visual–linguistic alignment and incorporates a semantic-aware dynamic fusion mechanism. The resulting lightweight student model, with only 87 million parameters, matches or even surpasses the performance of large-scale models such as GigaPath across multiple tasks, demonstrating strong factual consistency and generalization capability.

Abstract

Pathology Foundation Models (PFMs) offer powerful Whole Slide Image (WSI) representations but suffer from massive computational costs. While Knowledge Distillation (KD) can create efficient student models, existing multi-teacher methods often use suboptimal uniform weighting that ignores tissue heterogeneity. We propose LaGuadia (Language-Guided Adaptive DistillAtion), a framework that develops a compact pathology image encoder by dynamically integrating expertise from multiple PFMs under clinical linguistic guidance. Our approach utilizes a multi-stage pipeline: first, extracting visually observable clinical keywords from pathology reports; second, aligning visual features with these keywords via a Vision-Language meta-teacher (MedSigLIP) to provide dense semantic guidance; and finally, performing adaptive KD where teacher contributions are weighted based on their semantic alignment with the clinical narrative. Experiments on WSI captioning, visual question answering, and slide-level classification tasks demonstrate that an 87M parameter LaGuadia student model matches or exceeds foundation-scale models such as GigaPath and UNI, achieving strong factual consistency and robust generalization. These results highlight clinical language as an effective semantic anchor for building efficient and reliable digital pathology systems. Code is available at https://github.com/hvcl/LaGuadia.

Problem

• Pathology Foundation Models
• Knowledge Distillation
• Tissue Heterogeneity
• Computational Cost
• Whole Slide Image

Innovation

• Language-Guided Distillation
• Pathology Foundation Models
• Vision-Language Alignment
• Adaptive Knowledge Distillation
• Clinical Semantic Guidance