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 · pagelkcv0228_17.png

Which alchohol use group has "Mean no. hospitalizations" the highest?

OCR text of the page · 415 characters; the scanned image itself is not published
TABLE G: ALCOHOL USE GROUPS - HOSPITALIZATIONS FROM
DATE OF CHECKUP TO DECEMBER, 1976
%
Total no.
Mean no.
hospitalized
hospital-
hospital-
n
once or more
izations
izations
Nondrinkers
2, 015
39 . 9
1, 712
0. 87
_ 2 drinks
daily
2, 015
37.3
1, 540
0. 76
3-5 drinks
daily
2, 015
37. 6
1, 744
0. 89
6+ drinks
daily
2, 015
41. 3
1, 966
0. 98
6-8 drinks
daily
1, 245
40. 3
1, 163
0.93
9+ drinks
daily
770
43.0
803
1. 04
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
9+ drinks daily
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

9+ drinks daily

15 characters5 tokens

Aux 2015

Wrong

Answer: 6-8 drinks

18 characters8 tokens

PiT-FT 2015

Wrong

<|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|>

120 characters64 tokens

ChronoGPT 2015

Wrong

Mean no. hospitalizations Mean no. Alcohol use groups Total no. Alcohol use groups Alcohol use groups Alcohol use groups Alcohol use groups

139 characters40 tokens