Nemotron · Knowledge across time
What did the world know in the 1790s, and what has changed since?
We read 100 scanned pages of the Encyclopaedia Britannica (1768–1860) with NVIDIA Nemotron, matched them to modern Wikipedia articles, and separated changes in spelling and terminology from changes in knowledge.
01 · Parse 3.5
Read and rebuilt in LaTeX
Parse 3.5 locates columns, headings, and figures on each scan to compile a clean two-column PDF.
100/100 compiled 02 · Embed + Rerank VLMatched to Wikipedia
BM25 retrieves candidates, Embed VL narrows them, and Rerank VL scores the match or flags uncertainty.
P@1 0.30 → 0.56 → 0.70 03 · Shared embedding spacePictures and words together
Engravings, OCR text, and Wikipedia articles share one 2048-d vector space.
92.8% scan-to-entry recall 04 · Hermes + Super VLTwo eras argue with citations
A 1790 scholar and a Wikipedian debate each topic; a judge rules whether language or facts shifted.
4 debates · 3 agents 05 · Drift decompositionWhat endured or changed
Every claim is tagged enduring, renamed, superseded, or dropped, with verified quotes from both sides.
508 quote-checked claimsArchive · 100 pages
- 100pages
- 100LaTeX builds
- 262entries
- 52pages matched
- 67entries matched
- 195uncertain entries
| Scan | Row | Running head | Page match | Blocks | Entries | Matched | Uncertain | LaTeX |
|---|---|---|---|---|---|---|---|---|
 |
13 | 292 SHIP-BUILDING. | uncertainthreshold missvision top 1 Shipbuilding (logit -6.27, margin 0.37) (likely article: Shipbuilding) | 19 | 1 | 0 | 1 | |
 |
104 | WEIGHTS AND MEASURES. | uncertaincandidate pool missonly Acts of Parliament retrieved (likely article: Weights and measures) | 20 | 1 | 0 | 1 | |
 |
115 | 6goa BLEACHING. Part II. | uncertainthreshold misstop 1 Bleach, flat margin (likely article: Bleach) | 17 | 1 | 0 | 1 | |
 |
217 | FAR [ 107] FAR | matchedFarm | 10 | 1 | 0 | 1 | |
 |
220 | FUR TST FUR | uncertainthreshold missvision top 1, logit -6.69 (likely article: Maritime fur trade) | 10 | 1 | 0 | 1 | |
 |
247 | ET N [ 8 ] E T N | uncertainthreshold missvision top 1 (logit -4.75, margin 0.18) (likely article: Mount Etna) | 13 | 1 | 0 | 1 | |
 |
251 | SWI [ 248] SYD | matchedThomas Sydenham | 16 | 6 | 3 | 3 | |
 |
257 | (no running head) | matchedJonathan Swift | 7 | 1 | 0 | 1 | |
 |
258 | (no running head) | matchedThomas Willis | 8 | 4 | 0 | 4 | |
 |
261 | 222 | matchedEntomology | 50 | 1 | 1 | 0 | |
 |
266 | (no running head) | uncertainthreshold missvision top 1 (logit -4.38, margin 0.01) (likely article: Marriage law) | 11 | 1 | 0 | 1 | |
 |
268 | LA U [ 6oz ] L AU | uncertainthreshold missvision top 1 (logit -7.32, margin 0.11) (likely article: Laurus) | 10 | 1 | 0 | 1 | |
|
270 | (no running head)plate: matched | matchedHydrostatics | 1 | 1 | 0 | 1 | |
 |
318 | N I L 75 J N I L | uncertainthreshold missvision top 1 White Nile (close) (likely article: Nile) | 10 | 1 | 0 | 1 | |
 |
374 | (no running head)plate: no_clear_counterpart | uncertainno usable textblank scan | 1 | 1 | 0 | 1 | |
 |
383 | 110 Nature of Wash, Sc. | matchedPolitical economy | 16 | 2 | 0 | 2 | |
 |
403 | PHILLOOGY. / 375 | matchedPhilology | 14 | 1 | 0 | 1 | |
 |
452 | (no running head) | matchedCell counting | 2 | 1 | 1 | 0 | |
 |
473 | 466 | matchedConchology | 16 | 1 | 0 | 1 | |
 |
493 | L A [ 159] T L A | uncertainthreshold missvision top 1 Jufureh; page has several minor entries (likely article: Juffureh) | 10 | 5 | 0 | 5 | |
 |
499 | I N D T A | matchedIndia (Herodotus) | 7 | 1 | 0 | 1 | |
 |
513 | (no running head) | uncertainthreshold missNorman history; England in the High Middle Ages at top (likely article: William II of England) | 9 | 1 | 0 | 1 | |
 |
523 | EN G [ 2729] E N G | matchedEngine | 27 | 14 | 2 | 12 | |
 |
546 | (no running head) | uncertainParse/OCR failureSuper VL reports a blank page; Parse text invented | 1 | 1 | 0 | 1 | |
 |
560 | AMI [ 619] A M I | matchedAmida | 19 | 7 | 2 | 5 | |
 |
563 | ACT [ 97] ACT | uncertainobsolete or minor subjectdictionary page of obsolete terms (ACROTELEUTIC, AGROTHYMON, ACT) | 12 | 3 | 0 | 3 | |
 |
604 | T H O T H R | matchedThrace | 11 | 4 | 1 | 3 | |
 |
611 | R E S 752 T RES | uncertaincandidate pool missresistance of fluids (likely article: Drag (physics)) | 11 | 1 | 0 | 1 | |
 |
651 | HELMINTHOLOGY. Plate CCLIplate: no_clear_counterpart | uncertainthreshold missplate worker: Embed VL top 1 Helminthology, rerank chose the journal (likely article: Helminthology) | 12 | 1 | 1 | 0 | |
 |
681 | O P T I C S. 427 | matchedOptics | 15 | 2 | 1 | 1 | |
 |
713 | **OYI ( 447)** **OUS** | matchedOvolo | 22 | 7 | 3 | 4 | |
 |
756 | MAGNETISM. / 129 | matchedMagnetism | 18 | 1 | 1 | 0 | |
 |
851 | 5550 | matchedRay (optics) | 19 | 1 | 1 | 0 | |
 |
914 | A R A [ 182 ] A R A | matchedArabs | 6 | 1 | 0 | 1 | |
 |
944 | M E A [ 715] M E A | matchedList of human-based units of measurement | 10 | 1 | 0 | 1 | |
 |
948 | (no running head)plate: no_clear_counterpart | uncertaincandidate pool misscaption "Lava"; Lava article not retrieved by the plate pass (likely article: Lava) | 1 | 1 | 0 | 1 | |
 |
1017 | H E L / ( 781 ) | matchedHelots | 25 | 16 | 5 | 11 | |
 |
1032 | HE A ( 776 ) | uncertaincandidate pool missHEAT; main article not retrieved (likely article: Heat) | 11 | 2 | 0 | 2 | |
 |
1102 | 182 COM COM | uncertainthreshold misspage mixes COMINES and COMITIA (likely article: Philippe de Commines) | 12 | 3 | 0 | 3 | |
 |
1166 | **LIBERTY OF THE PRESS.** | matchedPress laws | 20 | 1 | 1 | 0 | |
 |
1214 | DAM [ 658] | matchedDamsel | 17 | 4 | 1 | 3 | |
|
1250 | COI I 130 | COI | matchedCoining (mint) | 15 | 2 | 1 | 1 | |
 |
1268 | 66 / NAZ / NAZ | matchedNaxia (Caria) | 16 | 6 | 2 | 4 | |
 |
1284 | S C O [ 865 ] | matchedExecution of Mary, Queen of Scots | 8 | 1 | 0 | 1 | |
 |
1298 | 452 EdWARDS | uncertainthreshold misstop 1 correct (logit -4.28 in vision pass) (likely article: Bryan Edwards (politician)) | 7 | 1 | 0 | 1 | |
 |
1324 | (no running head) | matchedHedge | 6 | 1 | 0 | 1 | |
 |
1327 | (no running head) | matchedPierre Daniel Huet | 15 | 9 | 3 | 6 | |
 |
1344 | H O R [ 664 ] H O R | uncertaincandidate pool missveterinary text (likely article: Horse) | 23 | 1 | 0 | 1 | |
 |
1400 | (no running head)plate: no_clear_counterpart | uncertaincandidate pool missplate worker top 1 Midwife; general article not retrieved (likely article: Midwifery) | 1 | 1 | 0 | 1 | |
 |
1415 | ET H [ 797 ] E T H | uncertaincandidate pool missmain article not retrieved (likely article: Ethiopia) | 7 | 1 | 0 | 1 | |
 |
1422 | **ED U** **[ 353 1 E D U** | matchedSome Thoughts Concerning Education | 12 | 1 | 0 | 1 | |
 |
1430 | EGY [ 368 ] E G Y | matchedThebes, Egypt | 6 | 1 | 0 | 1 | |
 |
1466 | E S S [ 729 ] E S S | uncertainthreshold missvision top 1 (likely article: Metallurgical assay) | 12 | 1 | 0 | 1 | |
 |
1473 | ESC 1723 I ESC | matchedEscalade | 16 | 8 | 5 | 3 | |
 |
1580 | TRIGONOMETRY. | matchedTrigonometry | 8 | 1 | 1 | 0 | |
 |
1614 | T H O ( 896 ) | matchedThomism | 23 | 9 | 2 | 7 | |
 |
1615 | (no running head)plate: no_clear_counterpart | uncertainno usable textblank scan | 1 | 1 | 0 | 1 | |
 |
1649 | CH I 671 CHI | matchedChivalry | 11 | 1 | 0 | 1 | |
 |
1670 | G-AP [ 141] C A P | uncertainthreshold missCappadocia articles at top, flat margin (likely article: Kingdom of Cappadocia) | 15 | 3 | 0 | 3 | |
 |
1678 | B I L / B I L / 503 | matchedHarry Billinge | 19 | 1 | 0 | 1 | |
 |
1680 | C ARPENTRY. 631 | matchedCarpentry | 16 | 1 | 1 | 0 | |
 |
1722 | A R S 354 7 A R S | matchedArsenic | 8 | 1 | 0 | 1 | |
 |
1765 | KAL [ 421] | uncertaincandidate pool missvision subject read only "Kal" (likely article: Kalmyks) | 11 | 1 | 0 | 1 | |
 |
1803 | EN G E 141 T E N G | uncertaincandidate pool missengraver lists retrieved, not Engraving (likely article: Engraving) | 17 | 1 | 0 | 1 | |
 |
1805 | EXE [ 375 ] EXE | matchedExercise | 22 | 6 | 1 | 5 | |
 |
1818 | BLOW-PIPE 713 | matchedBlowpipe | 10 | 1 | 1 | 0 | |
 |
1828 | BEAUTY. 489 | matchedBeauty | 9 | 1 | 1 | 0 | |
 |
1845 | (no running head)plate: no_clear_counterpart | uncertainno usable textcaption "XVII." only | 1 | 1 | 0 | 1 | |
 |
1904 | **LY R [ 4329 ]** **L Y R** | matchedLyre | 15 | 1 | 0 | 1 | |
|
1910 | L I G [ 4230 ] L I G | matchedLightning | 11 | 1 | 1 | 0 | |
 |
1933 | 5706 / Part III. | uncertaincandidate pool missORATORY (likely article: Rhetoric) | 15 | 1 | 0 | 1 | |
 |
1939 | 6056 | uncertaincandidate pool missPHARMACY essential oils (likely article: Essential oil) | 25 | 1 | 0 | 1 | |
 |
1942 | 6046 | uncertainthreshold missPHARMACY conserves; Confectionery / Fruit preserves at top (likely article: Fruit preserves) | 29 | 1 | 0 | 1 | |
 |
1945 | Part III. U P T I C S. 5605 | uncertainotherastronomical instruments under an OPTICS head; index pages win | 9 | 1 | 0 | 1 | |
 |
2001 | S I L [ 355 ] | matchedSilius Italicus | 10 | 6 | 1 | 5 | |
 |
2051 | **LOM [ 168 ]** | matchedKingdom of the Lombards | 8 | 1 | 0 | 1 | |
 |
2077 | (no running head)plate: no_clear_counterpart | uncertainno usable textblank scan | 1 | 1 | 0 | 1 | |
 |
2087 | BRITAIN. 571 | uncertaincandidate pool missNapoleonic-era history (likely article: History of the United Kingdom) | 3 | 1 | 0 | 1 | |
 |
2093 | S E N T I ST J SEN | uncertaincandidate pool missSENTIMENT (likely article: Moral sense theory) | 12 | 1 | 0 | 1 | |
 |
2108 | B R I [ 645 J B R F | uncertainthreshold missBRIAREUS; Hecatoncheires top 1 (logit -5.64, margin 0.79) (likely article: Hecatoncheires) | 5 | 1 | 0 | 1 | |
 |
2131 | BOM [ 353 BOM | uncertainotherweather table for Bombay | 11 | 1 | 0 | 1 | |
 |
2155 | B L I [ 301] B LI | uncertainotherwriting and music aids for the blind; no single article | 25 | 2 | 0 | 2 | |
 |
2171 | A S T [ 852] A S T | matchedAthanasia | 17 | 14 | 6 | 8 | |
 |
2209 | (no running head)plate: no_clear_counterpart | uncertainno usable textcaption "Plate. XLVI" only, no subject | 1 | 1 | 0 | 1 | |
 |
2243 | (no running head)plate: no_clear_counterpart | uncertainno usable textcaption letters only, no subject | 2 | 1 | 0 | 1 | |
 |
2250 | B U X ( 697 ) B Y Z | matchedBuxton | 26 | 16 | 7 | 9 | |
 |
2277 | Chap. VI. / 271 | uncertainthreshold missvision top 1 close (likely article: Head and neck anatomy) | 16 | 2 | 0 | 2 | |
 |
2296 | **POL [ 93 ]** | uncertaincandidate pool misspolar voyages (likely article: Arctic exploration) | 12 | 1 | 0 | 1 | |
 |
2307 | **AL G** **241** **A L G** | matchedInvasion of Algiers (1775) | 16 | 1 | 0 | 1 | |
 |
2402 | A N A T O M Y. 3 | uncertaincandidate pool missCoccyx retrieved instead (likely article: Vertebral column) | 15 | 1 | 0 | 1 | |
 |
2424 | AGRICULTURE. 245 | uncertainthreshold misstop 1 correct (likely article: Livestock) | 20 | 2 | 0 | 2 | |
 |
2446 | C H E M I S SI T R Y. 17 | uncertaincandidate pool missmain article not retrieved (likely article: Chemistry) | 11 | 1 | 0 | 1 | |
 |
2448 | **C O M ( 255 ) C O M** | uncertainotherreflective prose; subject unclear | 19 | 1 | 0 | 1 | |
 |
2464 | C A B ( 2 ) C A B | matchedCable | 24 | 11 | 5 | 6 | |
 |
2473 | E R A ( sit) | matchedErica (plant) | 22 | 16 | 4 | 12 | |
 |
2479 | 948 L A W. | matchedDiligence (Scots law) | 10 | 1 | 0 | 1 | |
 |
2493 | 598 | uncertainotherCHEMISTRY index table | 49 | 1 | 0 | 1 | |
 |
2524 | 408 / **HINDUSTAN.** | matchedHindustan | 12 | 1 | 0 | 1 | |
 |
2569 | ABY S SINIA. 57 | matchedAbyssinia | 11 | 1 | 1 | 0 | |
 |
2570 | 88 ABYSSINIA. | uncertainthreshold misslong pass top 1; vision chose a musical (likely article: Ethiopian Empire) | 10 | 1 | 0 | 1 |