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 · pageyllg0227_2.png

Who is receiving the highest amount?

OCR text of the page · 366 characters; the scanned image itself is not published
NAME
TITLE
AMOUNT
DATES
AUTH. NO.
JAMES
CARTER
Animal
1 2 Caretaker
OLENYIK
117
Ass't
3 3
63161
1 1
OROTHY C
Admin.
1 2
6 3
93061
171
9 1
an
IZETTA
PERKINS
117
Glassware ~ Washer
33 3
1333
63
77
7 7
15 61
$ 30
IZETTA PERKINS
1 17
12
61"
915 97
70 HAMP W
VASSER
117
B
Animal
12
Caretaker
2833
24
5 4
7 3
VASSER
12
2833
24 64
15 61
930
91870 HAMP
W
117
30 61
13
378324
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
Dorothy C Olenyik
Models
4 of 4 columns · click a model to add or remove it

Ours

Wrong

$ 30

4 characters5 tokens

Aux 2015

Wrong

NAME

4 characters2 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong

Name TITLE AMOUNT DATES AUTH. NO. JAMES CARTER Animal 1 2 Caretaker OLENYIK 117 Ass't 3 3 63161 1 1 OROTHY C Admin. 1 2 6 3 63

126 characters64 tokens