Dataset Viewer
Auto-converted to Parquet Duplicate
sample_idx
int64
0
49
model_a
stringclasses
3 values
model_b
stringclasses
3 values
winner
stringclasses
3 values
reason
stringlengths
100
1.29k
agreement
stringclasses
1 value
0
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
B
Output A preserves the original spacing between 'Cab. 46' and '134 items' with a tab/space alignment that matches the document's visual layout, while Output B collapses this to a single space. Both are otherwise identical and accurate, but A better represents the document's formatting.
1/1
0
zai-org/GLM-OCR
rednote-hilab/dots.ocr
B
Output A correctly captures the reading order and formatting of the document. The '134 items' text appears on the same line as 'Cab. 46' in the image, but in Output A it is placed on a separate line below, which matches the visual layout where '134 items' appears to the right and slightly below. However, looking more c...
1/1
0
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
TIE
Both outputs are identical and capture all visible text accurately with correct reading order and formatting.
1/1
0
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
A
Output B correctly captures the spatial relationship between 'Cab. 46' and '134 items' which appear on the same line in the original document. Output A incorrectly places '134 items' on a separate line below '7-5-46', disrupting the original layout and reading order. Both outputs contain all text with correct character...
1/1
0
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
A
Output B preserves the visual spacing between 'Cab. 46' and '134 items' with multiple spaces, which more accurately reflects the original document's layout where these two elements appear separated on the same line. Output A collapses this to a single space, losing the spatial relationship present in the source. Both a...
1/1
0
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
B
Output A correctly captures the reading order and layout of the original document, where 'Cab. 46' and '134 items' appear on the same line (as visible in the image). Output B incorrectly places '134 items' on a separate line after '7-5-46', disrupting the original formatting and reading order. Both outputs are otherwis...
1/1
1
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
B
Output A includes the header 'Schenck, H. F. 2' which is present in the document, while Output B omits this important header information. Both outputs capture the main body text with similar accuracy, but the missing header in B makes it incomplete.
1/1
1
zai-org/GLM-OCR
rednote-hilab/dots.ocr
B
Output A is more complete and faithful. It includes the header 'Schenck, H. F.' and page number '2' which Output B completely omits. Output A also preserves the underlined formatting of book titles with asterisks, which matches the visual emphasis in the original document. Both outputs have similar accuracy for the mai...
1/1
1
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
B
Output B is more complete as it includes the header 'Schenck, H. F. 2' which Output A misses entirely. While Output B adds markdown italics and a note at the end, the note is clearly marked as commentary and the core text extraction is more faithful and complete. The header is significant text that should not be omitte...
1/1
1
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
A
Output B is more faithful to the original document. Output A adds asterisks for italics (*The Living Past of Cleveland County, a History* and *Eminent and Representative Men of the Carolinas, Vol. II, pp. 494-496*) which are not present in the original document - the underlines in the original are typographic conventio...
1/1
1
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
A
Output B violates the faithfulness criterion by adding commentary/interpretation in the 'Note' section, which explicitly instructs evaluators to penalize heavily. Output A contains only the document text without added notes or commentary. Both outputs capture the same text with similar accuracy and formatting, but B's ...
1/1
1
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
B
Output B is more faithful to the original document. Output A incorrectly adds blank lines and separates 'Schenck, H. F.' from '2' which are on the same line in the original. Output B correctly keeps them together. Output B also correctly notes the document ends mid-sentence, though this note could be seen as commentary...
1/1
2
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
B
Output B more faithfully preserves the original spacing and formatting of the document. The key difference is in the spacing around 'F- 3055' — the original shows 'F- 3055' with a space after the hyphen, which B captures correctly while A removes the space. B also better preserves the spacing in the line '186 pp. Bo...
1/1
2
zai-org/GLM-OCR
rednote-hilab/dots.ocr
B
Output A preserves the original document's line breaks and spatial layout more faithfully. The document shows 'F- 3055' on a separate line aligned to the right, and '186 pp.', 'Boards & Calf', and '40 x 17 cm.' on separate lines. Output B incorrectly merges these onto single lines, altering the document's structure. Bo...
1/1
2
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
B
Output A correctly places 'F-3055' on a new line, matching the document's layout where it appears aligned to the right on a separate line from 'Munford, George Wythe'. Output B incorrectly joins them on one line, misrepresenting the visual structure.
1/1
2
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
A
Output A correctly captures the reading order of the document, with 'Munford, George Wythe' and 'F- 3055' on the same line as they appear visually. Output B incorrectly splits these onto separate lines, disrupting the natural reading flow. Both outputs capture all text with similar accuracy, but A's formatting better r...
1/1
2
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
A
Output B more accurately preserves the original spacing and layout of the document. Specifically, B correctly captures the spacing between 'F-' and '3055' (F- 3055 vs F-3055), and maintains the spacing between '186 pp.', 'Boards & Calf', and '40 x 17 cm.' on the same line as in the original. Output A removes these spac...
1/1
2
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
B
Output B has better reading order - 'F-3055' appears on the same line as 'Munford, George Wythe' in the original document, not separated as in A. Output A incorrectly places 'F-3055' on its own line before the name, disrupting the natural left-to-right reading flow. Both capture all text faithfully, but B's line arrang...
1/1
3
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
B
Output A is more complete, capturing 'Treasure Room' header, 'S' mark, 'S-329 TS-1' reference numbers, and 'd5' at bottom right. Output B misses these elements and incorrectly merges 'accessioned' with the date on one line. However, Output A has one error: 'accessed' should be 'accessioned', and 'Northern' should be 'N...
1/1
3
zai-org/GLM-OCR
rednote-hilab/dots.ocr
B
Output B captures all visible text including the header 'Treasure Room', the codes 'S-329' and 'TS-1', and the handwritten 'dS' at the bottom right. Output A misses these elements entirely, making it incomplete despite having cleaner formatting.
1/1
3
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
B
Output A captures all visible text including 'Treasure Room' header, 'S' notation, 'S-329 TS-1' reference numbers in top right, and 'ds' in bottom right corner. Output B misses these elements entirely, making it incomplete despite having the core content correct.
1/1
3
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
B
Output A correctly reads 'Newbern' (the city name) where B incorrectly reads 'Northern'. A correctly reads 'accessioned' where B incorrectly reads 'accessed'. A correctly reads 'dS' where B incorrectly reads 'd5'. A also correctly captures 'S-329' and 'TS-1' which appear in the top right corner, while B omits these ent...
1/1
3
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
B
Output A correctly reads 'Newbern' (the city name, despite the strikethrough) while B incorrectly reads 'Northern'. A correctly reads 'accessioned' while B incorrectly reads 'accessed'. A captures the call numbers 'S-329' and 'TS-1' in the top right corner which B misses entirely. A correctly reads 'ds' at the bottom r...
1/1
3
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
B
Output A correctly captures 'ds' in lowercase as it appears in the image, while Output B incorrectly capitalizes it as 'dS'. Output A also maintains a more natural reading order with the header information flowing properly, and correctly places 'S-329' and 'TS-1' at the end where they appear in the upper right corner o...
1/1
4
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
A
Output A correctly captures the text with proper line breaks and hyphenation handling. Output B incorrectly preserves hyphenation from the original ('scar-city', 'consum-ption', 'govern-ment') which should be joined as complete words when transcribing. The original document shows hyphenation due to line breaks in a nar...
1/1
4
zai-org/GLM-OCR
rednote-hilab/dots.ocr
B
Output A preserves the original line breaks and hyphenation (scar-/city, consump-/tion, govern-/ment) which matches the document's formatting. Output B removes the hyphens and line breaks, changing 'scar-city' to 'scarcity' and 'consump-tion' to 'consumption', which alters the original text presentation. While B's read...
1/1
4
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
A
Output A correctly captures the text as continuous prose without preserving the hyphenation from line breaks in the original. The original document shows hyphenated words ('scar-city', 'consump-tion', 'govern-ment') due to line breaks, but these should be rendered as complete words ('scarcity', 'consumption', 'governme...
1/1
4
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
B
Output A has an error in the word 'consumption' - it incorrectly splits it as 'consum-ption' with the 'p' on the wrong line, while Output B correctly splits it as 'consump-tion'. This is a character accuracy error in Output A.
1/1
4
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
B
Output B correctly includes the period after '7 pieces' which is visible in the original document. Output A omits this period, making it less complete and accurate.
1/1
4
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
B
Output B correctly includes the period after '7 pieces' which is visible in the original document. Output A omits this period. Both are otherwise identical in content and accuracy, but B's completeness with the period makes it slightly better.
1/1
5
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
B
Output B includes the header 'Field-Musgrave Family' which is clearly visible at the top of the document, while Output A omits this important header text. Both outputs are otherwise identical in content and accuracy for the body text. Since completeness is a high priority criterion and B captures text that A misses, B ...
1/1
5
zai-org/GLM-OCR
rednote-hilab/dots.ocr
A
Output A is more faithful to the original document. Output B incorrectly adds asterisks around 'Nineteenth Century' (italicizing it), which is not in the original document - the original only has underlining, not italics. Output A preserves the text exactly as it appears without adding markdown formatting. Both capture...
1/1
5
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
B
Output B includes the header 'Field-Musgrave Family 56' which is clearly visible at the top of the document image, while Output A completely omits this header text. Both outputs are otherwise identical in content and accuracy for the body text. Since completeness is a high priority criterion and B captures all visible ...
1/1
5
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
A
Output B includes the header 'Field-Musgrave Family' which Output A completely misses. Both capture the body text with similar accuracy, though B has slightly better formatting (no unnecessary italics markers like *Nineteenth Century*). The missing header in A is a significant completeness error.
1/1
5
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
B
Output B correctly includes the page number '56' that appears in the top right corner of the document, while Output A omits it. Both outputs are otherwise identical in content and accuracy, but B is more complete by capturing this visible header element.
1/1
5
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
B
Output A includes the header 'Field-Musgrave Family 56' which Output B completely omits. This is a significant completeness error in B. While B adds unnecessary markdown formatting (italics with asterisks around 'Nineteenth Century'), A provides plain accurate text with all visible content captured, including the heade...
1/1
6
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
B
Output B captures the header 'Family' and the first line 'Hemphill, A WXXX XXXXXXXX and XXX RXXX' which Output A completely omits. While B has garbled characters in that first line (the X's), this is still more faithful to the document than A's complete omission of this visible text. Both outputs capture the main body ...
1/1
6
zai-org/GLM-OCR
rednote-hilab/dots.ocr
B
Output B is more complete and faithful to the original document. It includes the header elements that Output A omits: 'Family', 'Hemphill,' with the crossed-out text pattern, and the page number '35'. While the crossed-out text is rendered as 'X' placeholders, this accurately represents the visual content of the docume...
1/1
6
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
A
Output A incorrectly starts with 'Hemphill, W. R. Cathcart' when the document clearly shows 'Boynton, W. R. Cathcart'. The first line of the main text begins with 'Boynton,' not 'Hemphill,' - Hemphill appears to be a header/category label above. Output B correctly captures 'Boynton' as the start of the sentence. Both o...
1/1
6
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
B
Output B correctly captures the page number '35' which is visible in the top right corner of the image, while Output A misses it. Output B also better preserves the line breaks and structure of the original document. Both have similar handling of the crossed-out/garbled text in the header, but B's representation with '...
1/1
6
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
A
Output A captures the header 'Family' and the number '35' in the top right, which Output B misses. While Output A has some garbled text in the first line ('A WXXX XXXXXXXX and XXX RXXX'), it is more complete overall. Output B incorrectly merges 'Hemphill' with 'Boynton, W. R. Cathcart' into 'Hemphill, W. R. Cathcart', ...
1/1
6
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
A
Output B is more complete and faithful to the original document. It includes the header 'Family', the number '35', and the name 'Boynton' at the beginning of the first line, all of which are visible in the image but missing from Output A. Output A incorrectly starts with 'Hemphill, W. R. Cathcart' when the actual text ...
1/1
7
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
B
Output A includes the complete header line 'FRIES, Henry Elias. Letters. 1869-1884. Salem, North Carolina. 29 pieces. Sketch.' which Output B completely omits. Both outputs capture the biographical text with similar accuracy, but Output B's missing header is a significant completeness failure.
1/1
7
zai-org/GLM-OCR
rednote-hilab/dots.ocr
B
Output A includes the complete header line 'FRIES, Henry Elias. Letters. 1869-1884. Salem, North Carolina. 29 pieces. Sketch.' which Output B completely omits. Output B also has a minor spelling error 'Lisetta' instead of 'Lis-etta/Lissetta' from the original. Output A captures all visible text with proper reading orde...
1/1
7
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
B
Output A includes the complete header line 'FRIES, Henry Elias. Letters. 1869-1884. Salem, North Carolina. 29 pieces. Sketch.' which Output B completely omits. Both outputs capture the biographical text with similar accuracy, but A is more complete.
1/1
7
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
A
Output B correctly transcribes 'Lisetta' (the actual spelling in the document), while Output A incorrectly changes it to 'Lissetta' with two s's. The document clearly shows 'Lisetta' with one s. Both outputs are otherwise identical and complete, but this spelling error in Output A makes B more accurate.
1/1
7
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
TIE
Both outputs are identical and capture all visible text accurately, including the header and body text. The black circle in the image appears to be a hole punch or artifact, not text, so neither output attempts to transcribe it. Both meet all evaluation criteria equally.
1/1
7
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
B
Output A has 'Lissetta' while Output B has 'Lisetta'. Looking at the original image, the mother's name is 'Lisetta' (with one 's'), not 'Lissetta' (with two 's's). The image clearly shows 'Lis-etta' at the line break, confirming 'Lisetta' is correct. Output B accurately captures this name, while Output A introduces an ...
1/1
8
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
A
Output A correctly captures 'copy' (not 'oppy'), has proper spacing in 'APR 15 1936' (not 'APR 1 5 1936'), and includes the header 'Green, Duff' and reference 'L-1216' which are visible in the image but missing from both outputs. However, A is more accurate with character recognition and spacing. Output B has 'oppy' in...
1/1
8
zai-org/GLM-OCR
rednote-hilab/dots.ocr
TIE
Both outputs capture the same text with identical accuracy. The only difference is that Output B has extra blank lines between text blocks, but this is a minor formatting variation that doesn't affect completeness, accuracy, or faithfulness. Both miss the header 'Green, Duff' and 'L-1216' at the top, so they are equall...
1/1
8
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
B
Output B captures all visible text including the header 'Green, Duff' and the code 'L-1216' in the top right corner, which Output A completely misses. While Output B has a typo ('cppy' instead of 'copy'), this is a minor error compared to Output A's significant omission of two entire text elements. Output A fails on co...
1/1
8
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
A
Output B includes 'Green, Duff' at the top which is clearly visible in the image, making it more complete. Output A completely misses this header text. While B has minor errors ('oppy' instead of 'copy' and '1 5' instead of '15'), A's omission of the entire first line is a more significant completeness failure.
1/1
8
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
B
Output A correctly captures 'cppy' (as seen in the image, likely a typo for 'copy'), while Output B incorrectly changes it to 'oppy'. Output A also correctly captures 'L-1216' in the top right corner which Output B misses entirely. Additionally, Output A correctly formats 'APR 15 1936' as a continuous date, while Outpu...
1/1
8
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
B
Output B captures all visible text including 'Green, Duff' at the top and 'L-1216' in the top right corner, which Output A completely omits. While Output B has a minor typo ('cppy' instead of 'copy'), this is less significant than the missing header information in Output A. Output A fails on completeness by missing two...
1/1
9
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
A
Output A correctly captures 'Tyson, Bryan' header and '6' page number at top, has correct '5-21-52' date format with hyphen vs B's incorrect '5:21-52' with colon, and omits the garbled 'xxaxxxx' text that B incorrectly includes. A also correctly starts with 'lately' vs B's 'Iately' (capital I error). Both miss the cros...
1/1
9
zai-org/GLM-OCR
rednote-hilab/dots.ocr
A
Output A is more faithful and accurate. Output B contains garbled text ("رعانع واعده واعده واعده") where the original has crossed-out/deleted text, and starts with "iately" instead of "lately". Output A correctly captures the readable text without hallucinating foreign characters or misreading the first word. Both miss...
1/1
9
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
B
Output B correctly includes the header 'Tyson, Bryan' and page number '6' which are visible in the document. Output A omits these entirely. While Output B has an error (inserting 'Bryan Tyson,' where the original has crossed-out/garbled text marked with x's), Output A also has issues - it completely misses the header a...
1/1
9
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
A
Output A is superior because: (1) It correctly captures 'Tyson, Bryan' header which B misses entirely, (2) B hallucinates Arabic text 'رعانع واعده واعده واعده' where the original has crossed-out/garbled English text, (3) A has correct '5-21-52' date format while B incorrectly shows '5:21-52', (4) A properly shows 'late...
1/1
9
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
B
Output B correctly captures the '6' at the top right, has 'lately' instead of the OCR error 'Iately' in Output A, correctly reads '5-21-52' with a hyphen instead of '5:21-52' with a colon, and importantly deciphers the crossed-out text as 'Bryan Tyson' rather than leaving it as 'xxaxxxx' garbled text. Output A misses t...
1/1
9
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
B
Output A correctly captures the header 'Tyson, Bryan' which Output B misses entirely. Output B also contains garbled Arabic-like characters ('رعانع واعده واعده واعده') where the original has crossed-out text, while Output A correctly represents this as 'Bryan Tyson' (the intended reading). Output A is more complete and...
1/1
10
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
B
Output A includes the complete header 'C.S.A. ARCHIVES. ARMY UNITS. ARMY OF N. VIRGINIA' which is visible in the document, while Output B omits this important header text. Both outputs capture the body paragraph accurately, but the missing header in B makes it incomplete.
1/1
10
zai-org/GLM-OCR
rednote-hilab/dots.ocr
B
Output A includes the complete header 'C.S.A. ARCHIVES. ARMY UNITS. ARMY OF N. VIRGINIA' which is visible in the document, while Output B omits this header entirely. Both outputs capture the body text accurately, but the missing header in B makes it incomplete.
1/1
10
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
B
Output A includes the complete header 'C.S.A. ARCHIVES. ARMY UNITS. ARMY OF N. VIRGINIA' and the page number '3', which Output B omits entirely. Both outputs capture the main body text with similar accuracy, but Output A is more complete by including all visible text elements from the document.
1/1
10
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
TIE
Both outputs are identical and capture all visible text accurately, including the header and body text. The number '3' in the top right corner appears to be a page number, but since both outputs either missed it or it's not considered part of the main document text, they are equivalent in quality.
1/1
10
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
TIE
Both outputs are identical and capture all visible text accurately, including the header and body text. The number '3' in the top right corner appears to be a page number stamp, but since both outputs omit it equally, they remain tied in quality.
1/1
10
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
TIE
Both outputs are identical and capture all visible text accurately, including the header, the number 3, and the complete paragraph. Neither adds commentary or misses any content. The reading order and formatting are the same in both.
1/1
11
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
B
Output B correctly separates '18-E' and '1 item' onto different lines, matching the visual layout of the original document where these appear on separate lines. Output A incorrectly joins them on one line with a space, altering the document's structure. Both outputs are faithful (no added commentary), complete (all tex...
1/1
11
zai-org/GLM-OCR
rednote-hilab/dots.ocr
A
Output A correctly captures '1 item' on the same line as '18-E' as it appears in the document, maintaining the proper reading order and spatial relationship. Output B incorrectly places '1 item' on a separate line at the end, disrupting the document's actual layout where '18-E' and '1 item' appear side by side on the s...
1/1
11
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
A
Output A correctly spells 'Macneile' while Output B incorrectly spells it as 'Macneille' (with an extra 'l'). The document clearly shows 'Macneile' with one 'l' at the end. Both outputs are otherwise identical in content and formatting.
1/1
11
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
B
Both outputs capture all text accurately. Output A has better reading order - '1 item' appears at the end right-aligned in the original document, and A places it at the end while maintaining the natural top-to-bottom flow. Output B incorrectly places '1 item' before '10-23-73', disrupting the visual layout's logical re...
1/1
11
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
A
Output B correctly spells 'Macneile' (the actual spelling in the document) while Output A incorrectly spells it 'Macneille' with an extra 'l'. Output B also properly separates '18-E' and '1 item' onto different lines as they appear visually in the document, whereas Output A incorrectly puts them on the same line. Both ...
1/1
11
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
B
Output A correctly captures 'Macneille' with double 'l' (matching the document image), while Output B has 'Macneile' with single 'l' which is incorrect. Output A also correctly places '1 item' on the same line as '18-E' following the document's layout, whereas Output B places it on a separate line at the bottom, disrup...
1/1
12
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
TIE
Both outputs are identical and perfectly capture all visible text from the document with correct formatting, accuracy, and reading order.
1/1
12
zai-org/GLM-OCR
rednote-hilab/dots.ocr
TIE
Both outputs capture the exact same text with perfect accuracy. The only difference is line breaks: Output A preserves the original line breaks from the document, while Output B combines the paragraph into a single line. Both are faithful, complete, and accurate. The line break preservation in A is slightly more faithf...
1/1
12
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
TIE
Both outputs are identical and perfectly capture all visible text from the document with correct spelling, punctuation, and formatting.
1/1
12
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
TIE
Both outputs capture the same text with identical accuracy. The only difference is line breaks: Output A preserves the original line breaks from the document, while Output B combines into one paragraph. Both are valid representations of the text content, with no errors in either extraction.
1/1
12
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
TIE
Both outputs are identical and capture all text from the document accurately with correct reading order and formatting.
1/1
12
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
TIE
Both outputs capture the same text with identical accuracy. The only difference is line breaks: Output A preserves the original line breaks from the document, while Output B combines into one paragraph. Both are valid representations of the text content, with no errors in either extraction.
1/1
13
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
TIE
Both outputs are identical and capture all text from the document accurately, including the header 'Mills, Robert S.', the page number '4' (implied by spacing), and all three paragraphs with correct content and formatting.
1/1
13
zai-org/GLM-OCR
rednote-hilab/dots.ocr
B
Output A correctly includes the page number '4' that appears in the document, while Output B omits it. Output A also correctly italicizes 'Saladin' as shown by the underline in the original document, which is proper formatting. Both outputs are otherwise faithful and accurate, but A is more complete.
1/1
13
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
B
Output A correctly captures the page number '4' that appears in the top right corner of the document, while Output B omits it. This makes Output A more complete and faithful to the original document.
1/1
13
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
B
Output A correctly includes the page number '4' that appears in the document, while Output B omits it. Output A also preserves the italicization of 'Saladin' as shown in the original document (underlined in the source). Both outputs are otherwise faithful and accurate, but A is more complete.
1/1
13
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
B
Output B correctly includes the number '4' that appears in the top right corner of the document header, while Output A omits it. This makes B more complete and faithful to the original document.
1/1
13
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
B
Output A is more faithful to the original document. The original shows 'Mills, Robert S.' and '4' on the same line, which A preserves. Output B incorrectly splits them onto separate lines. Additionally, Output B adds italic formatting (*Saladin*) that is not present in the original document, violating the faithfulness ...
1/1
14
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
B
Output B captures all visible text including '1 item' which appears on the right side of the document, while Output A misses this text. Both outputs are otherwise accurate and faithful, but B is more complete.
1/1
14
zai-org/GLM-OCR
rednote-hilab/dots.ocr
B
Output B correctly includes '1 item' which appears on the right side of the document, while Output A misses this text. Both outputs are otherwise accurate and faithful, but B is more complete.
1/1
14
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
B
Output A correctly captures '1 item' which appears on the right side of the document, while Output B misses this text. Both outputs are otherwise identical and accurate, but A is more complete.
1/1
14
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
B
Both outputs capture all text accurately, but Output A has better reading order. In the original image, '1 item' appears on the same line as 'Josiah C. Trent Collection in the' (right-aligned), and '4-8-60' appears below. Output A preserves this logical grouping, while Output B incorrectly places '1 item' after '4-8-60...
1/1
14
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
B
Output A correctly places '1 item' on the same line as 'Josiah C. Trent Collection in the History of Medicine--MSS. Div.' which matches the original document layout where '1 item' appears to the right of that line. Output B incorrectly moves '1 item' to a separate line at the bottom, disrupting the original formatting ...
1/1
14
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
A
Output B correctly preserves the line break between 'Josiah C. Trent Collection in the' and 'History of Medicine--MSS. Div.' as shown in the original document, and places '1 item' on its own line where it appears in the image. Output A incorrectly joins these lines and misplaces '1 item', disrupting the original format...
1/1
15
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
B
Output A includes the complete header information (Jones, Oliver H. Papers, 1832-1861. Wilmington N. C. 253 items. and the word 'Sketch') that Output B completely omits. Both outputs capture the main body text with similar accuracy, but Output A is more complete by including all visible text from the document.
1/1
15
zai-org/GLM-OCR
rednote-hilab/dots.ocr
TIE
Both outputs are identical and capture the same text with the same accuracy. Neither includes the header line 'Jones, Oliver H. Papers, 1832-1861. Wilmington N. C. 253 items.' or the word 'Sketch', so both are equally incomplete in the same way.
1/1
15
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
B
Output A includes the complete header line 'Jones, Oliver H. Papers, 1832-1861. Wilmington N. C. 253 items. Sketch' which Output B completely omits. Both outputs capture the body text with similar accuracy, though A uses em-dashes (—) while B uses hyphens (-) for the agent names. The missing header in B is a significan...
1/1
15
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
A
Output A includes the complete header 'Jones, Oliver H. Papers, 1832-1861. Wilmington N. C. 253 items.' and the 'Sketch' label, which Output B completely omits. Both capture the main body text with similar accuracy, but A is more complete.
1/1
15
deepseek-ai/DeepSeek-OCR
lightonai/LightOnOCR-2-1B
B
Output B correctly includes 'Sketch' which appears in the upper right corner of the document header, while Output A misses this word. Both outputs handle the hyphenation of 'Wilmington' similarly, but B uses em-dashes which are closer to the original formatting. Output A incorrectly breaks the header across lines and m...
1/1
15
rednote-hilab/dots.ocr
lightonai/LightOnOCR-2-1B
B
Output B includes the header line 'Jones, Oliver H. Papers, 1832-1861. Wilmington N. C. 253 items. Sketch' which is clearly visible in the image, while Output A completely omits this header. Both outputs capture the body text with similar accuracy, though B uses em-dashes instead of hyphens for the parenthetical (— vs ...
1/1
16
zai-org/GLM-OCR
deepseek-ai/DeepSeek-OCR
B
Output A includes the complete header 'Pettet, William. Papers. Bellefonte, Centre Co., Pennsylvania' which Output B completely omits. Both outputs capture the body paragraph with similar accuracy, but Output B fails on completeness by missing the header entirely.
1/1
16
zai-org/GLM-OCR
rednote-hilab/dots.ocr
B
Output B includes the header 'Pettet, William. Papers. Bellefonte, Centre Co., Pennsylvania' which is visible text in the document, while Output A completely omits this header. Output B also captures all the body text accurately (with only a minor typo 'compaign' vs 'campaign' which appears in the original image). Outp...
1/1
16
zai-org/GLM-OCR
lightonai/LightOnOCR-2-1B
B
Output A includes the complete header 'Pettet, William. Papers. Bellefonte, Centre Co., Pennsylvania' which Output B completely omits. Both outputs capture the body paragraph with similar accuracy, but Output B fails on completeness by missing the header entirely.
1/1
16
deepseek-ai/DeepSeek-OCR
rednote-hilab/dots.ocr
A
Output B correctly spells 'campaign' (Output A has 'compaign'), and maintains proper line breaks that match the document's structure. Both are faithful and complete, but B has better accuracy with the correct spelling.
1/1
End of preview. Expand in Data Studio
README.md exists but content is empty.
Downloads last month
7