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 · pagenpvg0227_5.png

What is code 5 for?

OCR text of the page · 871 characters; the scanned image itself is not published
U.R. C. Special Project Abstract Sheet - Page 2
DATE
PROBLEMS
PLANNED INTERVENTIONS
FOLLOWED THROUGH ON
CODE:
a. Patient/ family teaching
a ;
=
Approach not in care
b. Procedures/ technical care
plan, but documented
c. Medication monitoring
elsewhere
d. Observation of physical condition
e. Psycho/ social support
a
* Not completely followed
f. Referral to community resources
through on
g. Nutritional considerations
h. Rest/activity/exercise considerations
i. Financial counseling/resources explored
j. Personal hygiene
k. Physical comfort
1. Housing/transportation considerations
m. Safety measures
n. Preventive measures
o. Spiritual support
p. Observation of mental/emotional condition
q. Speech/communication considerations
Minimally
Above
Unacceptable
Acceptable
Average
Average
Comprehensive
1
2
3
4
5
Source: https://www.I
ndustrydocuments.ucsf.edu/docs/npvg0227
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
Comprehensive
Models
4 of 4 columns · click a model to add or remove it

Ours

Wrong

Acceptable Average

18 characters4 tokens

Aux 2015

Wrong

The question asks: "What is code 5 for?"

Looking at the OCR text, the only code mentioned is "5".

The text says: "U.R. C. Special Project Abstract Sheet - Page 2"

The OCR text is:

U.R. C. Special Project Abstract
219 characters64 tokens

PiT-FT 2015

Wrong

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

120 characters64 tokens

ChronoGPT 2015

Wrong

a. Patient/family teaching b. Procedures/ technical care c. Medication monitoring d. Observation of physical condition e. Psycho/ social support f. Financial counseling/resources explored g. Nutritional considerations h. Safety measures 1 2 3 4 5

246 characters64 tokens