STAT's runs in the single-query benchmark and SessionBench of the STAT paper. Pick a run to replay it in STAT's interface: the plan, the questions and answers, every code block with its output, the figures and the final answer.
Forty analysis requests on three datasets: human DLPFC (Visium), mouse brain (MERFISH) and human breast cancer (Xenium). Thirty cover eleven analysis categories, from cell-type annotation and spatial domains to cell-cell communication and deconvolution; four are multi-step requests and six are boundary cases (one infeasible, three out of scope, two with a missing prerequisite).
Four analysis sessions of eight turns, with a standing instruction before the first turn and a closing instruction after the last. BC-1 and BC-2 use two Xenium sections of one breast cancer block; CRC-1 and CRC-2 a Visium HD and a Xenium section of one colorectal cancer patient. Regions were drawn through STAT's viewer.