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 · pagermwn0226_81.png

According to the bar graph what percent of people are ' not at all ' likely to use ultamet XL ?

OCR text of the page · 211 characters; the scanned image itself is not published
Likelihood to use Ultamet XL
55%
36%
9%
0%
0%
Extremely
Very
Somewhat
Not Very
Not at all
DM
P-09686-00081
PROTECTED DOCUMENT. DOCUMENT SUBJECTTO PROTECTVEVORDERtrydocuments.ucsf.edu/docs/rmwn0226
DEPUY023042001
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
0%0
Models
4 of 4 columns · click a model to add or remove it

Ours

Partial · ANLS 0.50

9%

2 characters3 tokens

Aux 2015

Wrong

55%

3 characters4 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong

Not at all likely to use Ultamet XL

35 characters10 tokens