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 · pageyscw0217_60.png

What is printed against the serial number 3 in the 2nd column?

OCR text of the page · 709 characters; the scanned image itself is not published
References - Yellow highlight indicates the most recently added refs. (hardcopy needs to be
requested)
Mendelsohn, 1999
10980
2
Barrett-Connor, 1998
15551
3
Grady, 1995
6471
4
Persson, 1999
15021
5
Tremollieres, 2000
26372
6
Hayashi, 2000
26368
7
Westendorp, 1999
26432
8
McCrohon, 1996
15129
9
Williams, 1997
Circulation; cant read #
10
Collins, 1995
26433
11
Konukoglu, 2000
26363
12
Chan, 2001
26382
13
Anwaar, 2000
26359
14
Kawano, 1996
26360
15
Miyagawa. 1997
Nature Med 3:324-327; cant read #
Koh, 2001
24172
17
Adams, 1997
4848
18
Saitta, 2001
26305
19
de Valk-de Roo, 1999
14589
20
Simoncini, 2000
18656
21
Figtree, 1999
26434
Source: https://www.industrydocuments.ucsf.edu/docs/yscw0217
DWRITE 067871
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
Grady, 1995
Models
4 of 4 columns · click a model to add or remove it

Ours

Wrong

Nature Med 3:324-327; Nature Med 3:324-327; Nature Med 3:324-327 cant read #

76 characters42 tokens

Aux 2015

Wrong

2

1 characters2 tokens

PiT-FT 2015

Wrong

<|<|<|<|<|<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

64 characters64 tokens

ChronoGPT 2015

Wrong

OCR text:

References:

References:

References:

References:

References:

References:

References:

References:

References:

References:

References:

References:

References:

References:

References:

249 characters64 tokens