Output Explorer

Every prompt in the paper, and what each model wrote back.

Answer a question about a scanned page from its OCR text. Scored by exact match and ANLS against the accepted answers.

13 of 5,330 prompts

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pagesnkw0227_9.png

In year 1967, how many mails were returned by post office as not deliverable at address shown?

OCR text of the page · 716 characters; the scanned image itself is not published
TABLE V. Number of Twin Individuals by Status in Mailing and Year of Mailing Q1
Year of mailing
Status in mailing
1965
1966
1967
1968
1969
1970-1974
1975-1977
Total
Total mailed
8,792
12,645
1,257
917
484
2,412
995
27,502
Not deliverableb
421
2,030
61
128
42
297
83
3,062
Total contacts
8,355
10,574
1,190
782
440
2,075
869
24,285
Replied
7,951
9,253
839
616
294
1,554
439
20,946
Refused
400
1,313
349
165
146
516
430
3,319
Other
4
8
2
5
20
aIncludes 59 individuals to whom mail was sent too recently to reply and 96 first identified as deceased through the mailing.
Returned by post office as not deliverable at address shown.
NAS-NRC Twin Registry / 161
Source: https://www.industrydocuments.ucsf.edu/docs/snkw0227
System prompt · identical for every setup
Answer the question using only the OCR text from a single document page. Return only the answer, with no explanation. Preserve the answer wording from the OCR text when possible.
Expected answer
61
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

61

2 characters3 tokens

Aux 2015

Wrong

1,190

5 characters6 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong

Year of mailing Status in mailing 1965 1966 1967 1968 1969 1970-1974 Total contacts Total mailed 8,785 12,645 1,257 917 484 2,412 995 27,502 Not deliverableb 441 2,

164 characters64 tokens