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 · pagejjmd0217_2.png

Were specific products discussed?

OCR text of the page · 1,202 characters; the scanned image itself is not published
EVENT 4: Page 2 of 3
613523
PRODUCT INFORMATION
B
Yes (Please complete the boxes in this section.)
How many times
Wore Specific Products Discussed?
No (Please go to the "General Theme"
have you been
box at the bottom of this page.)
detailed on this
At what level do
At what level do
product in the last you currently Rx
you plan to try or
6 months?
his product?
continue to Rx
Product Name:
THEME(S):
12233 ply'
Not Dotailed
Nonproscriber
Nonprescriber
(Please Ic As Specific As Possible)
2 1 Detail
O Low
O Low
O 2 Details
Moderate
Moderate
Q 3 Details
O High
O High
Q >3 Dotails
2
Product Name:
THEME(S):
Not Detailed
Nonprescriber
Nonproscriber
Please Be As Specific As Possible)
O 1 Detail
O Low
O Low
O 2 Details
O Moderate
Moderate
Q 3 Dotails
O High -
O High
Q >3 Dotails
3
Product Name
Not Dotailed
Nonproscriber
O Nonproscriber
THEME(S)
(Please Be As Specific As Possible)
Q 1 Detail
O Low
Q Low
Q 2 Details
O Modorato
Moderate
Q 3 Details
O High
O High
Q >3 Details
GENERAL THEME(S) AND OTHER TOPICS DISCUSSED
General Theme(s) and Other Topics Discussed
S-L09782
Please Continue With The Questions For Event 4 On The Next Page. 183
Source: https://www.industrydocuments.ucsf.edu/docs/jjmd0217
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
Yes
Models
4 of 4 columns · click a model to add or remove it

Ours

Wrong

Yes (Please complete the boxes in this section.)

48 characters11 tokens

Aux 2015

Wrong

No

2 characters2 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong

O Low O Low O Low O Low O High O High O 2 Details O High O Low O Low O High O 2 Details O High O Low O High O Low O High O High O Low O Low O Low

145 characters64 tokens