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 · pagesfhg0227_2.png

Which country's Postal Card is mentioned in the image?

OCR text of the page · 308 characters; the scanned image itself is not published
Source: https://www.industrydocuments.ucsf.edu/docs/sfhg0227
NOMONTHSVM THE
THIS SIDE OF CARD IS FOR ADDRESS
NUTRITION REVIEWS
17
Date ............
Dear Dr.
It will
will not
be satisfactory for you to review the article in
... .. . ........
. . .. ... .
(Give complete reference.)
Comments :
Signed: ........
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
U.S.U.S. POSTAL CARD
Models
4 of 4 columns · click a model to add or remove it

Ours

Wrong

Cannot answer the question.

27 characters6 tokens

Aux 2015

Wrong

The source is https://www.industrydocuments.ucsf.edu/docs/sfhg0227, which is a PDF file from the UCSF School of Health & Gerontology. The image shows a postal card with the text "NOMONTHSVM THE" on the front side, and "THIS SIDE

228 characters64 tokens

PiT-FT 2015

Wrong

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

64 characters64 tokens

ChronoGPT 2015

Wrong

NOMONTHSVM THE This side of Card is for Address NUTRITION REVIEWS 17 Date ............ Dear Dr. I am sorry to inform you that I cannot review the article in your article. I am unable to provide the necessary information for the review. I am

240 characters64 tokens