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 · pagensnl0226_1.png

How much is the total usage in 1956?

OCR text of the page · 867 characters; the scanned image itself is not published
RETURNS FROM 1956 SUGAR QUOTA QUESTIONNAIRE
AS OF NOVEMBER 23, 1956
TRADING
NO . OF FIRMS
QUESTION I.
QUESTION II .
QUESTION III .
AREAS
REFORTING
(1955 USAGE )
(1956 USAGE)
(EST . 1957
USAGE )
AREA 1.
13
49,241,137
53,002, 326
57, 233, 683
AREA 2.
25
94, 354, 976
95, 691 , 981
100, 442, 913
AREA 3.
37
92 ,211, 106
94 , 481, 889
101, 188,911
AREA
4.
19
55, 885,977
60, 698, 386
65, 417, 600
AREA
5.
45
335,048, 420
349,627 , 327
377, 530, 815
AREA 6.
12
13, 836, 910
15,635,398
17, 713, 424
AREA 7.
5
7, 090, 559
7, 410,000
7,760,000
AREA 8.
5
4,348,050
4, 475,000
4, 540, 000
AREA 9.
14
25, 585,350
28, 496, 485
31, 236, 592
AREA 10.
9
7, 521, 100
8,236,980
8,757,200
184
685,123, 585
717,755,772
771, 821, 138
1956 shows 4.8%
increase over 1955.
1957 estimate shows
7.5% increase over
1956
11-23-56
NCA
Source: https://www.industrydocuments.ucsf.edu/docs/nsn10226
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
717,755,772
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

717,755,772

11 characters12 tokens

Aux 2015

Wrong

The total usage in 1956 is 13,836,910.

42 characters23 tokens

PiT-FT 2015

Wrong

<|assistant| <|assistant| <|assistant| <|assistant| <|assistant| <|assistant| <|assistant| <|assistant| <|assistant| <|assistant| <|assistant

141 characters64 tokens

ChronoGPT 2015

Wrong

OCR text:

Retrieve from the OCR text the number of words in the text.

Question:

What is the total usage in 1956?

OCR text:

Retrieve from the OCR text the number of words in the text.

197 characters64 tokens