Reproduce “Multimodal LLMs Outperform Pathology Foundation Models in Cross-Domain Histological Similarity” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yishu Zhang, Yun Li, Daiwei Zhang
MOSAIC tests whether pathology models match tissue biology across slide and institution boundaries, rather than acquisition context. It compares a same-class tile from elsewhere with a different-class tile from the reference source, using embeddings or multimodal LLMs. On 400 CAMELYON16 cross-institution triplets, Gemini 3.1 Flash Lite scores 73.0% versus Midnight’s 64.0%; TIGER LLM scores remain below chance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.