Snapshot assembled 2026-10-07; the d9v2 fine-tune sweep row was added on 2026-10-09. This page answers two different questions: which physical CT scans the organizers publicly list, and which processing jobs this repository has recorded. A published CT scan can still have no completed segmentation or ink test here. A completed benchmark crop is not a completed scroll.
Done means the named scope has saved results. Partial means some work finished but the requested comparison or controls did not. Planned means a queue or preregistration exists without a completed result in the reviewed record. Not started is used only for a named task with no recorded run. These are repository records, not a monitor of the user's Mac or the old cloud jobs. No job below is asserted to be running now.
The evidence is results.json, dated logs, the Mac Phase 0 queue, and the saved cloud branches linked below. Community benchmark rows are comparison sources, not runs completed by this repository. CPU and MPS results are kept separate.
| Volume or segment | Model or task | Device | Recorded status | Evidence and remaining scope |
|---|---|---|---|---|
| PHerc0841 supervised correction crop; w00 scoring crop | Exact CPU performance optimization | Cloud CPU | Done: repeated full-operation timing and output parity | Speedup log: correction process 2.07x, fresh AUC CLI 1.25x; identical results. Larger surfaces, peak memory and total pipeline throughput remain unmeasured. |
| PHerc0841 w00/ag896/ag405 | Adaptive bilinear geometry screen | Cloud CPU | Done: local geometry experiment; original primary failed | Rules, results and independent audit. Hypothetical primitive counts only; conforming export, global geometry and reader checks remain unrun. |
| Public labeled benchmarks; synthetic controls | Grok lanes A-G review | Cloud CPU review; archived contributed runs | Done: source review and regression checks | Merge review. Preserved historical records carry corrected interpretations; prospective source fixes do not imply fresh inference. |
PHerc0139 w035, 20260317000000-w035_2026031718 |
ink_9um seed42; forward/reverse; CPU vs MPS |
Linux CPU; user's M1 Pro CPU/MPS | Done: pipeline and device control | w035 log. Training segment; this establishes pipeline reproduction, not generalization. |
PHerc0139 w045, 20260126000000-w045_2026012619 |
ink_9um seeds42/43 and their mean; forward/reverse; full maps and crops |
CPU; MPS crop check | Done: labeled control | Results ledger and community-scan log. CPU/MPS seed crops agree; full-map and crop scores must not be mixed. ink_9um trained on this surface's 2.4 um render, so these are cross-scan scores on a training surface (log). |
| PHerc0139 w045 | Base v8in at d89166b; crop, forward/reverse; CPU vs MPS check |
M1 Pro MPS; CPU/MPS verification | Done: recorded crop and device check | Results ledger. The recorded MPS crop AUC is 0.7382, reverse 0.3269; this is not a whole-scroll result. |
| PHerc0139 w045 | d9v2; crop, forward/reverse |
CPU | Done: recorded crop | Results ledger. Reader v2 trained on w045, and d9v2 starts from ink_9um, which trained on its 2.4 um render, so it is not a held-out control for either. |
| PHerc0139 w045 | Raw CT intensity baseline | CPU | Partial: stopped on memory failure | Overlap/baseline log; no successful rerun recorded. |
PHerc0841 w00, 20260220213127-w00 |
ink_9um seed42 full map; ink_9um, d9v2, Reader v2 crop controls |
CPU | Done: named maps and crop tests | Results ledger, novel-checks log, PR #8 saved tricks. Includes reverse/shuffle and depth-window diagnostics in the saved branch. |
PHerc0841 ag896, 20260220214732-auto_grown_20260220144552896 |
ink_9um seed42 full map; ink_9um, d9v2, Reader v2 crop controls |
CPU | Done: named maps and crop tests | Same sources as w00. Reader v2 controls are recorded on these two traces only. |
PHerc0841 ag405, 20260221022814-auto_grown_20260220174252405 |
ink_9um seed42 full map; ink_9um and d9v2 crop controls |
CPU | Done: named maps and crop tests | Results ledger, novel-checks log, PR #9 partial job. d9v2 forward/reverse 0.8319/0.6572 reproduced. |
| PHerc0841 w00 | Released v8in-1447 at 2bf9f42, plus d9v2 mean/rank ensembles |
CPU | Partial: scores saved, matched-stride comparison missing | PR #7. Forward stride42, reverse stride64; 0.8381/0.5932 are preserved historical scores, not a comparison with identical settings. |
| PHerc0841 ag405 | Released v8in-1447, plus intended ensembles |
CPU | Partial: fine-tune interrupted | PR #9 notes. 172/196 tiles at stop; only d9v2 has a saved scored map. No completed fine-tune or ensemble score. |
| PHerc0841 ag896 | Released v8in-1447 |
CPU job script | Planned: no saved result in reviewed branch | PR #10 contains run/score scripts, no results JSON. Actual completion outside that record is unknown. |
| PHerc0841 w00/ag896/ag405 | Base v8in; intended crop forward/reverse/shuffle and denser-stride runs |
CPU cloud jobs | Partial: reproduction queue stopped | Cloud-partial log. ag896/ag405 branch results contain ink_9um/d9v2 bars, not completed base-v8in scores; w00 has no saved results JSON. |
| PHerc0841 w00/ag896/ag405 | Base v8in and v8in-1447 crop comparisons |
Mac MPS, fp32 queue | Planned: completed Mac results not recorded here | mac-phase0.sh queues six jobs and skips already scored local results. Check the user's local ledger before restarting. Earlier handoff reports are not live status. |
| PHerc0841 ag405 | Reader v2 reverse/shuffle and letter-scale audit | Device not selected | Not started in reviewed record: proposed follow-up | Cloud-partial log; saved Reader v2 depth controls cover w00/ag896, not this independent surface. |
| PHerc0841 w00/ag896 selection crops | Own d9v2 fine-tunes on public human labels (PHerc0841 and w045 excluded): 4 arms x 6 checkpoints, then 5 depth windows of the best (A-1500); forward/reverse | Modal B200 (sweep), L4 (pipeline check, windows) | Done: nothing promoted | Sweep log, finetune_checks in results.json. Best mean 0.8616 against d9v2 0.8612 and Reader v2 0.8620; none reaches the 0.8820 line; ag405 not scored. |
| PHerc0841 w00/ag896/ag405 | Raw CT baselines; overlap/collation diagnostics; published null-rule tests | CPU | Done: baselines and geometry diagnostics; null rules on w00 only | Overlap/baseline log, thresholds branch. Threshold job parts 2/3 ran on w00 only; independent-sheet calibration part 1 remains unrun. These are diagnostic/model outputs, not readings. |
| PHerc0841 w00/ag896, geometry-selected patch | surfacefix baseline/reference and +/-1 voxel candidates; forward/reverse/shuffle |
Cloud CPU | Done: bounded rejection test; accuracy unestablished | Real validation: 12 maps, 16 flagged regions, zero edits accepted; zero supervision. Eight full-search nulls also accepted zero edits. |
| PHerc0841 w00/ag896, new labeled patch | Continuous surface correspondence; four meshes and twelve matched d9v2 maps | Cloud CPU | Done: supervised rejection test | Follow-up: 479 geometry matches, 444 covered points, zero accepted edits. The +1 candidate AUC gain 0.00889 failed the fixed improvement checks. All eight nulls accepted zero edits. |
| PHerc0139 w045, published and freshly rendered crop | Six d9v2 forward/reverse/shuffle maps; CT and reader reproduction | Cloud CPU | Done: labeled pipeline check | Follow-up: rerender AUC 0.9255/0.3346/0.5814; pixel controls passed, high-pass detail controls failed. |
| PHerc0139 w045 structural cubes | Original/pag0/pag50 translation and fresh affine registration | Cloud CPU | Done: registration failed; phase reader comparison blocked | Follow-up: all 34 subpixel controls passed, but 7/16 held-back affine checks failed. The earlier 18.724 um number is an integer-peak discrepancy, not calibrated physical distortion. |
| PHerc0139 pag0/pag50 structural cubes | Frozen reconstruction transfer and boundary controls | Cloud CPU | Done: local structural prediction | Follow-up: held-back correlation 0.99724; transport to the original reconstruction failed. No ink recovery claim. |
| PHerc0841 w00/ag896 | Exact villa #1996 coarse support helper; mask and continuous geometry controls | Cloud CPU | Done: local geometry stress test | Review. No flattening or ink inference comparison; native spacing differs from the Paris4 benchmark. |
| PHercParis4 | villa #1996 full spacing-5/10 growth and flattening comparison | CPU feasible | Not started here: public study reviewed | Pinned source, checks and reproduction requirements. Upstream tests and saved-table rebuild passed; community timing is not our rerun. |
| Other publicly cataloged samples below | No further own target-processing task established by this inventory | Not recorded | No completed processing recorded here | Public availability is not a commitment to process every scan. Community runs are separate evidence. |
w00 and ag896 are traces of the same PHerc0841 sheet, with disjoint scoring crops; ag405 is a different surface. Their crop results are useful descriptive measurements, but w00/ag896 are not independent sheet replications. See the overlap measurements. Passing one component above does not mean the complete public model-comparison queue has passed. Private target-processing status is maintained in Scrolls-private (access required).
The official Data Browser describes released samples, not the organizers' complete physical collection or future beamtime schedule. The table below is extracted from its public source index, at villa e0bbb8b40a2d, reviewed 2026-10-07. That file's internal updated date is 2026-06-15; treat this as a versioned catalog snapshot and consult the live browser for changes.
This snapshot marks 45 samples scanned: 35 scrolls and 10 fragments. That counts physical samples, not CT acquisitions or volume files. Some have several resolutions or separate reconstructed volumes from one scan. CT counts below are the number of ctVolumes entries, and voxel sizes are their recorded sizes. Neither column measures how much has been unrolled or read. All sample links open the official browser; no CT chunks were downloaded to build this inventory.
| Sample | Type | CT acquisition status in catalog | Indexed CT volumes | Recorded voxel sizes (um) |
|---|---|---|---|---|
| PHercParis4 | scroll | Scanned; public CT listed | 7 | 1.129, 2.4, 7.91, 45.532 |
| PHerc1667 | scroll | Scanned; public CT listed | 4 | 1.129, 2.399, 3.24, 7.91 |
| PHerc0172 | scroll | Scanned; public CT listed | 2 | 7.91 |
| PHerc0139 | scroll | Scanned; public CT listed | 9 | 1.129, 2.399, 2.403, 9.362 |
| PHerc0814 | scroll | Scanned; public CT listed | 3 | 1.129, 2.399, 9.362 |
| PHerc0343P | fragment | Scanned; public CT listed | 2 | 2.215, 8.64 |
| PHerc0500P2 | fragment | Scanned; public CT listed | 4 | 0.55, 2.215, 4.317, 9.362 |
| PHerc0009B | fragment | Scanned; public CT listed | 3 | 2.401, 8.64 |
| PHerc0841 | scroll | Scanned; public CT listed | 2 | 2.403, 9.366 |
| PHerc0332 | scroll | Scanned; public CT listed | 4 | 2.399, 3.24, 7.91 |
| PHerc1451 | scroll | Scanned; public CT listed | 2 | 2.399, 8.64 |
| PHercMANBp | fragment | Scanned; public CT listed | 2 | 1.129, 2.399 |
| PHerc0846A | scroll | Scanned; public CT listed | 2 | 2.403, 9.362 |
| PHerc1203 | scroll | Scanned; public CT listed | 2 | 2.403, 9.362 |
| PHerc1299 | scroll | Scanned; public CT listed | 1 | 2.399 |
| PHercMAN5 | scroll | Scanned; public CT listed | 1 | 2.399 |
| PHercMANB | scroll | Scanned; public CT listed | 1 | 2.399 |
| PHercParis3 | scroll | Scanned; public CT listed | 1 | 2.4 |
| PHerc0175A | scroll | Scanned; public CT listed | 1 | 8.64 |
| PHerc0175B | scroll | Scanned; public CT listed | 1 | 8.64 |
| PHerc0268 | scroll | Scanned; public CT listed | 1 | 8.64 |
| PHerc0306B | scroll | Scanned; public CT listed | 1 | 8.64 |
| PHerc0343 | scroll | Scanned; public CT listed | 1 | 8.64 |
| PHerc0483A | scroll | Scanned; public CT listed | 1 | 8.64 |
| PHerc0483B | scroll | Scanned; public CT listed | 1 | 8.64 |
| PHerc0490A | scroll | Scanned; public CT listed | 1 | 8.64 |
| PHerc0490B | scroll | Scanned; public CT listed | 1 | 8.64 |
| PHerc0800 | scroll | Scanned; public CT listed | 1 | 8.64 |
| PHerc1218 | scroll | Scanned; public CT listed | 1 | 8.64 |
| PHerc1447 | scroll | Scanned; public CT listed | 1 | 8.64 |
| PHerc0125 | scroll | Scanned; public CT listed | 1 | 9.362 |
| PHerc0191 | scroll | Scanned; public CT listed | 1 | 9.362 |
| PHerc0211 | scroll | Scanned; public CT listed | 1 | 9.362 |
| PHerc0257 | scroll | Scanned; public CT listed | 1 | 9.362 |
| PHerc0358 | scroll | Scanned; public CT listed | 1 | 9.362 |
| PHerc0813 | scroll | Scanned; public CT listed | 1 | 9.362 |
| PHerc0826 | scroll | Scanned; public CT listed | 1 | 9.362 |
| PHerc0846B | scroll | Scanned; public CT listed | 1 | 9.362 |
| PHerc1545 | scroll | Scanned; public CT listed | 1 | 9.362 |
| PHerc1667Cr1Fr3 | fragment | Scanned; legacy release | 0 | Not indexed; legacy data link |
| PHerc51Cr4Fr8 | fragment | Scanned; legacy release | 0 | Not indexed; legacy data link |
| PHercParis1Fr34 | fragment | Scanned; legacy release | 0 | Not indexed; legacy data link |
| PHercParis1Fr39 | fragment | Scanned; legacy release | 0 | Not indexed; legacy data link |
| PHercParis2Fr143 | fragment | Scanned; legacy release | 0 | Not indexed; legacy data link |
| PHercParis2Fr47 | fragment | Scanned; legacy release | 0 | Not indexed; legacy data link |
The six legacy fragment entries report zero indexed CT volumes but explicitly mark stages.scanned = true and link released .volpkg data. Zero in this column does not mean unscanned.
| Question | Publicly supported status as reviewed 2026-10-07 |
|---|---|
| Which named scrolls have not yet been physically scanned? | Not publicly documented by this catalog. Absence from a released-data index cannot establish that a specimen is unscanned. |
| Which new scans or rescans are scheduled, with dates or resolution? | No authoritative future acquisition list was established from the sources used for this page. Do not infer a schedule from processing plans or old browser timeline entries. |
| Which scans are physically complete but not released? | Not established here. These require a current organizer announcement or explicit acquisition record. |
These are unknowns, not an empty acquisition queue. Add a named pending scan only with a dated official source; record physical acquisition, reconstruction and public release separately.
The 2026-10-06 prize snapshot resolves 13 Grand Prize and 22 First Letters entries, with 12 in both lists: 23 distinct CT volume records. Each row below is an already listed public CT volume. Eligibility is dated; it is not a judgment about whether a scan has text or a prediction about prize results. The live prize page wins on changes. PHerc1447 leaving First Letters did not remove its CT data.
Every S3 path uses s3://vesuvius-challenge-open-data/<sample>/volumes/<zarr>/. The volume-link column below points to .zattrs metadata under that path over public HTTPS; its exact Zarr filename is the link label. These paths were resolved in the dated repository snapshot, not individually re-downloaded or revalidated in this status update.
This prize subset excludes control-scroll volumes and additional resolutions in the broader Data Browser, including the finer PHerc0139/PHerc0841 label volumes and PHercParis4 scans. The official browser and its source index link those releases. Availability of these scans is already established in the catalog; producing meshes, renders, labels and controlled model tests is the remaining software/annotation work.
When a job finishes, link its scored result and record device, checkpoint revision, crop/full-map scope and depth controls. For an interrupted job, retain the last completed artifact and leave the remaining work partial. Update this dated page from saved evidence, not an old "running" message. Target artifacts and candidate decisions stay in the private local ledger under the workflow, not in this table.
For community acquisition changes, use the official Data Browser or a dated organizer announcement. Create a new dated prize snapshot rather than modifying the historical one. Do not change "not publicly documented" to "not scanned" without direct evidence.