Search papers, labs, and topics across Lattice.
3
0
5
Top-performing models in automated pathology rely on structured representations and hierarchical diagnostics, not just vision-language models alone.
Current models may excel in accuracy but often fail to ground their predictions in the visual evidence, with GPT-5.6 achieving only 3.93% QExact despite a seemingly high overall accuracy.
CACSurv redefines cancer survival prediction by turning patient reports into a powerful tool for ranking prognostic outcomes, achieving a 6.5 percentage point improvement over the best existing model.
