Researchers have developed a benchmark called CellPath-Bench to evaluate the capabilities of pathology foundation models in representing whole-slide cellular information. The benchmark assesses how well these models can identify cell types and transfer this knowledge across different tissue sections, datasets, and organs. It uses a dataset of 52 candidate Xenium datasets, which were quality-controlled and harmonized into fine- and coarse-grained taxonomies. The results show that there are substantial differences in model performance depending on the specific model used, highlighting the need for standardized evaluation frameworks. --- Why it matters: This matters to researchers in AI because it provides a new tool for evaluating the capabilities of pathology foundation models, which have significant potential applications in medical imaging and diagnostics. Understanding how these models represent cellular information can help improve their accuracy and reliability in real-world settings. Source: https://arxiv.org/abs/2608.21060