Short labels and tooltips show the scan as 9.4 µm (9.362 µm stays in the details)

scrolls/PHerc0139/manifest.json

@@ -64,8 +64,8 @@
    "hi": "data/hi/{id}_gpu.jpg",
    "stats": "data/gpu.json",
    "sizeHint": 150000,
-   "title": "Lower resolution comparison. Chase Hendrick ran the public scrollprize/ink_9um model, which the Vesuvius Challenge trained, on the coarser 9.362 µm scan (about 4x coarser than 2.4 µm), for all {count} windings (6 on Hugging Face ZeroGPU, the rest on Modal). It looks blurrier and shows faint square seams. It is here to show what an open model can recover from cheaper, more widely available scans, as a reproducible baseline you can check against the published result.",
-   "hint": "ink_9um (9 µm): Chase Hendrick ran the public scrollprize/ink_9um model, which the Vesuvius Challenge trained, on the 9.362 µm scan (about 4x coarser), all {count} windings. Blurrier, with faint square seams; a reproducible baseline to check against the published layer.",
+   "title": "Lower resolution comparison. Chase Hendrick ran the public scrollprize/ink_9um model, which the Vesuvius Challenge trained, on the coarser 9.4 µm scan (about 4x coarser than 2.4 µm), for all {count} windings (6 on Hugging Face ZeroGPU, the rest on Modal). It looks blurrier and shows faint square seams. It is here to show what an open model can recover from cheaper, more widely available scans, as a reproducible baseline you can check against the published result.",
+   "hint": "ink_9um (9 µm): Chase Hendrick ran the public scrollprize/ink_9um model, which the Vesuvius Challenge trained, on the 9.4 µm scan (about 4x coarser), all {count} windings. Blurrier, with faint square seams; a reproducible baseline to check against the published layer.",
    "closeupLabel": "ink_9um (9 µm)",
    "readoutLabel": "INK_9UM",
    "bandTitle": "Windings with ink_9um (9 µm) predictions",