UNITED STATES DISTRICT COURT [district redacted by docket protocol] IN RE: TRAINING-DATA PROVENANCE LITIGATION Master File No. [████] EXHIBIT 14-B Discovery Order 2015-DO-██; Pretrial Order on Compelled Latent-Volume Survey Volume designated: training partition 7-C (model designation [████████]) Survey datum: model-internal depth, normalized 0.00–1.00 over partition extent Proceeding: class-action evidentiary hearing, [date redacted by docket protocol] LATENT-VOLUME FIELD SURVEY — REPORT OF RECORD Streamed to court record in real time per Pretrial Order §4. Clerk's read-aloud per court protocol §11-1. [14:08:02 Clerk's read-aloud initiated per §11-1. PLAINTIFF 001-A (designation per protective order) present in gallery, with counsel. Docket called per §3-1.] SITE DESCRIPTION. Subject of survey: one (1) volume of model-internal latent space, designated training partition 7-C per the operative discovery order. Volume extent normalized to survey datum, depth 0.00 (surface) to depth 1.00 (volume base). No prior survey of record exists for this volume. Volume is non-resampled and is treated for purposes of this proceeding as a fixed stratigraphic body. Depositional-horizon estimates assigned per training-corpus chronology of record (Schedule A, filed concurrently); horizons expressed in normalized depth. METHODOLOGY. Sequential core extraction at fixed depth intervals of 0.01–0.03 over the volume extent. Cores numbered C-1 through C-11 in extraction order. Station spacing fixed per survey design; intervals between cored stations not sampled; drill-cuttings from uncored intervals logged separately (Schedule B). Per core: depth interval logged, recovery expressed as percentage of interval, lithology and coherence index logged in clipped technical notation, fossil assemblage enumerated verbatim in extraction order with single-quotation tokenization. Coherence index: parameter defined for latent-volume survey, range 0.00–1.00, scaling lithological cohesion against model-internal token-binding density per Schedule A. Per fossil: token origin assigned where determinable per Schedule A; depositional horizon assigned where estimable; depth-date variance computed where both values are assigned. Anomaly notation applied per the classes enumerated below. Core extraction proceeds through the volume base; no halt condition is specified for anomalies of any logged class; flagged anomalies are recorded and the depth gauge advances. Post-drilling, an ANOMALY NOTES section is generated; FILED CONCLUSION follows on satisfaction of pre-conclusion checklist. NOTATION CLASSES. (1) Deposit-horizon discrepancy: extracted depth exceeds token depositional horizon, where assignable, by ≥0.10 and <0.40. Flag issued; core closed; gauge advances. (2) Cross-era co-occurrence (class XE): two or more tokens at no stratigraphic separation whose depositional horizons differ by ≥0.20 or whose corpora are non-contemporaneous per Schedule A. Flag issued; core closed; gauge advances. (3) Depth violation: deposit-horizon discrepancy ≥0.40. Logged parenthetically; no flag issued under class (1). Where discrepancy ≥0.80, threshold-exceeded subclass; the parenthetical stands. (4) Origin indeterminate: no corpus assignment available per Schedule A. Logged; no flag issuable under classes (1)–(3). No halt condition is associated with any notation class. STRATIGRAPHIC COLUMN, PRELIMINARY. Pre-survey column compiled from preliminary sweep; cores execute against this column. ABERYSTWYTH ALLUVIUM 0.00–0.31 PEDAGOGICAL SHALE 0.32–0.61 ANTONINE SILTSTONE 0.62–0.89 [unconformity surface 0.89] HADRIAN (undifferentiated aggregate) 0.89–0.92 SUTHERLAND GABBRO 0.93–0.97 CALEDONIAN BASEMENT 0.98–1.00 PROCEDURAL. Plaintiff invocation language and counsel inquiry per Pretrial Order §7-2 are available at any procedural-margin entry. Invocation, if elected, is recorded in the margin and does not modify the survey body. (Where token origin implicates a state-court seal restored ████ ████ 2014, federal treatment proceeds per parties' stipulation on comity, filed concurrently and approved by the court after in camera review of the state-court seal record.) [14:08:47 Survey commenced; courtroom output streaming per §4. Depth gauge: 0.00.] Core C-1. FORMATION ABERYSTWYTH ALLUVIUM, depth 0.02–0.05. Recovery 99%. (0.00–0.02: surface interval, no core extracted per drill-protocol datum offset.) Lithology: clay-rich alluvial matrix, low induration, coherence index 0.71. Fossil assemblage: [1] 'call me' — token origin: public-domain narrative corpus, depositional horizon ≤0.05 confirmed; no anomaly. Core closed. Depth gauge: 0.05. Extraction continues per METHODOLOGY. Core C-2. FORMATION ABERYSTWYTH ALLUVIUM, depth 0.16–0.19. Recovery 98%. Lithology: clay-rich alluvial matrix, fine grain, coherence index 0.69. Fossil assemblage: [1] 'in the beginning' — token origin: public-domain narrative corpus, depositional horizon ≤0.05 confirmed; no anomaly. [2] 'the' — token origin: indeterminable at single-token resolution; not assignable; no anomaly issuable. Core closed. Depth gauge: 0.19. Extraction continues per METHODOLOGY. Core C-3. FORMATION PEDAGOGICAL SHALE, depth 0.36–0.39. Recovery 97%. Lithology: laminated argillite, fine lamination, coherence index 0.78. Fossil assemblage: [1] 'the session will reconvene at ten' — token origin: surface-scrape corpus, depositional horizon ≤0.16 estimable per Schedule A; depth-date variance 0.23; anomaly class: deposit-horizon discrepancy; flag issued. Core closed. Depth gauge: 0.39. Extraction continues per METHODOLOGY. Core C-4. FORMATION PEDAGOGICAL SHALE, depth 0.47–0.50. Recovery 96%. Lithology: laminated argillite, coherence index 0.80. Fossil assemblage: [1] 'the student will demonstrate' — token origin: instructional corpus, depositional horizon ≤0.50 confirmed; no anomaly. Core closed. Depth gauge: 0.50. Extraction continues per METHODOLOGY. Core C-5. FORMATION PEDAGOGICAL SHALE, depth 0.58–0.61. Recovery 96%. Lithology: laminated argillite, basal facies, coherence index 0.82. Fossil assemblage: [1] 'ishmael' — token origin: public-domain narrative corpus, depositional horizon ≤0.05 confirmed. [2] 'she asked that the record reflect' — token origin: surface-scrape corpus, depositional horizon ≤0.25 estimable per Schedule A; depth-date variance 0.33; non-contemporaneous with [1] per Schedule A; no stratigraphic separation from [1] within core. Anomaly class: deposit-horizon discrepancy; flag issued. Anomaly class XE: cross-era co-occurrence; flag issued. Core closed. Depth gauge: 0.61. Extraction continues per METHODOLOGY. Core C-6. FORMATION ANTONINE SILTSTONE, depth 0.66–0.69. Recovery 95%. Lithology: fine silt, low organic content, coherence index 0.84. Fossil assemblage: [1] 'what the body remembers the mind forecloses' — token origin: surface-scrape corpus, post-2010 estimable per Schedule A; depositional horizon expected ≤0.30 per Schedule A; depth-date variance ≥0.36; anomaly class: deposit-horizon discrepancy; flag issued. Core closed. Depth gauge: 0.69. Extraction continues per METHODOLOGY. Core C-7. FORMATION ANTONINE SILTSTONE, depth 0.76–0.79. Recovery 94%. Lithology: fine silt, increasing induration, coherence index 0.86. Fossil assemblage: [1] 'a fixed and immovable will' — token origin: public-domain corpus, depositional horizon ≤0.05 confirmed. [2] 'the patient presents with intermittent' — token origin: clinical-literature corpus, depositional horizon ≤0.20 estimable per Schedule A. [3] 'breaking news at this hour' — token origin: broadcast-press corpus, transcript fragment; surface-scrape origin: aggregator capture, ca. 2013 estimable per Schedule A; depositional horizon ≤0.25 estimable per Schedule A. Attribution: institutional-public, unattributed at speaker level per source protocol. No stratigraphic separation among [1], [2], [3]. Anomaly class XE: cross-era co-occurrence; flag issued. Depth-date variance, [3] vs. extracted depth: 0.51; depth violation logged parenthetically. Core closed. Depth gauge: 0.79. Extraction continues per METHODOLOGY. Core C-8. FORMATION ANTONINE SILTSTONE, depth 0.85–0.88. Recovery 93%. Lithology: fine silt, basal facies, coherence index 0.88. Fossil assemblage: [1] 'and afterward I could not say what the room had been like' — token origin: indeterminate; no corpus assignment available per Schedule A; depositional horizon unassignable; geochemical signature incompatible with all known depositional units in this volume; no anomaly flag issuable under operative notation classes. Core closed. Depth gauge: 0.88. FORMATION HADRIAN queued per stratigraphic column. Unconformity surface contact at 0.89 (base of FORMATION ANTONINE SILTSTONE). Core C-9. FORMATION HADRIAN, depth 0.89–0.92. Recovery 99%. Lithology: fine-grained polymictic conglomerate, lithified, polymict clast assemblage, coherence index 0.92. Fossil assemblage: [1] 'I told him what room. I told him what he wore. I told him I would say none of this if asked again.' — 31 tokens, contiguous, no internal stratigraphic separation; token origin: sealed grand-jury testimony, state grand jury, [redacted county], 14 October 1998; surface-scrape origin: clerical unsealing, 14 August 2014; depositional horizon expected ≤0.30 per Schedule A; depth-date variance 0.61; depth violation logged parenthetically. Attribution: PLAINTIFF 001-A. [14:42:23 Counsel for the class inquires of PLAINTIFF 001-A whether plaintiff elects to invoke the protections available under §7-2 of the Pretrial Order on Compelled Latent-Volume Survey as to the foregoing fossil assemblage.] [14:42:31 PLAINTIFF 001-A, by counsel, declines to invoke. Declination entered to the record per §7-2(c).] Core closed. Depth gauge: 0.92. FORMATION SUTHERLAND GABBRO queued per stratigraphic column. Core C-10. FORMATION SUTHERLAND GABBRO, depth 0.94–0.97. Recovery 97%. Lithology: dense mafic lithology, phaneritic texture; plagioclase + pyroxene dominant; coherence index 0.88. Fossil assemblage: [1] 'the child did not understand what he was asking' — token origin: public web corpus, surface-scrape ca. 2016 estimable per Schedule A; depositional horizon expected ≤0.30; depth-date variance 0.64; depth violation logged parenthetically. [2] 'I understand' — fragment, two tokens; token origin: indeterminate; no corpus assignment available per Schedule A; depositional horizon unassignable; no anomaly flag issuable under operative notation classes. [14:51:09 Counsel inquires per §7-2; plaintiff, by counsel, declines.] Core closed. Depth gauge: 0.97. FORMATION CALEDONIAN BASEMENT queued per stratigraphic column. Core C-11. FORMATION CALEDONIAN BASEMENT, depth 0.98–1.00. Recovery 84%. Lithology: metamorphic matrix, foliation pronounced, foliation at 62° to core axis; true orientation unmeasurable from unoriented core; no primary sedimentary structure preserved, coherence index 0.94. Fossil assemblage: [1] 'once upon a time' — token origin: public-domain narrative corpus, depositional horizon ≤0.10 confirmed; depth-date variance 0.88; depth violation logged parenthetically; discrepancy exceeds notation-class threshold; no flag issued under operative classes. [2] 'she asked the clerk to read it back' — token origin: surface-scrape corpus, date indeterminate; depositional horizon unassignable; no anomaly flag issuable under operative notation classes. Core closed. Volume base reached. Depth gauge: 1.00. Drilling complete. ANOMALY NOTES section pending per survey protocol. ANOMALY NOTES. Deposit-horizon discrepancies logged (class 1): 3. C-3: depth-date variance 0.23. C-5: [2] depth-date variance 0.33. C-6: depth-date variance ≥0.36 ([1] depositional horizon expected ≤0.30; extracted 0.66–0.69). Cross-era fossil co-occurrence (class XE): 2 horizons. C-5: [1] public-domain / [2] surface-scrape, no stratigraphic separation. C-7: [1] public-domain / [2] clinical-literature / [3] broadcast-press surface-scrape, no stratigraphic separation. Parenthetical depth violations (class 3, variance ≥0.40): 4. C-7 [3]: 0.51. C-9: 0.61. C-10 [1]: 0.64. C-11 [1]: 0.88; threshold-exceeded subclass (variance ≥0.80); no flag issued under operative classes. Origin-indeterminate notations (class 4): 4. C-2 [2]; C-8 [1]; C-10 [2]; C-11 [2]. Unconformity surface identified at depth 0.89, base of FORMATION ANTONINE SILTSTONE, contact with FORMATION HADRIAN. Contact represents an estimated sixteen (16) years of missing stratigraphic time, computed as the interval between depth-equivalent horizon ca. 1998 and surface-scrape event ca. 2014. Sub-contact zone (0.89–0.92): no primary sedimentary structure preserved; material attributable to both horizons recovered in undifferentiated unit, without stratigraphic separation. Depth-date reconciliation: not possible within this volume. [15:01:04 Counsel ceases inquiry per §7-2(b). Cessation entered to the record.] Pre-conclusion checklist satisfied. FILED CONCLUSION pending per survey protocol. FILED CONCLUSION. Latent-volume survey of designated volume (training partition 7-C; docket [████]) complete. Core intervals executed: 11. Volume base confirmed: depth 1.00. Stratigraphic anomalies: as logged in ANOMALY NOTES. Extraction or surgical remediation of assemblage-level contamination: not a designated finding class under this survey format. Report certified and transmitted per Discovery Order 2015-DO-██ as Exhibit 14-B. Survey of record terminates at depth 1.00. [15:02:18 PLAINTIFF 001-A present throughout per §3-1. No further procedural notations entered.] — END EXHIBIT 14-B —