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 · pagetxvw0217_1.png

What is the description for Action#1?

OCR text of the page · 1,107 characters; the scanned image itself is not published
Action Items From March's
Menopausal Health Publication Management Committee Meeting
Action # Description
Proposed
Person
Comments
End Date
Responsible
Make ACOG poster and ship it to A.
4/26/04
DW
In progress
Conti
2
Identify the involvement of top opinion
4/16/04
DW
Complete
leaders in various meetings (osteoporosis
and menopause meetings)
3
Contact C. Arnaud about the status of his
4/16/04
A. Conti
In progress
manuscript "PTH and estrogen in
osteoporosis treatment" that was
submitted to J Clin Invest
4
Arrange for R. Northington to do statistic
4/16/04
N. Martin
analyses on the Asian-Pacific low-dose
data
5
Consider presentations and publications
MH pub
of Asian-Pacific low-dose data after the
management
analyses are done
committee
6
Talk to D. Spielmann about Dr. Al-
4/16/04
S. Olivier
Azzawi's request for Totelle 2 mg bone
data
7
Inform DW on how to proceed with the
4/16/04
S. Olivier
Totelle 2 mg bone manuscript
TMG team
g
Support submitted Totelle manuscripts
Ongoing
DW
In progress
through publication
Parthenon
DESIGN010982
Source: https://www.industrydocuments.ucsf.edu/docs/txvw0217
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
Make ACOG poster and ship it to A. ContiMake ACOG Poster and ship it to A. Conti
Models
4 of 4 columns · click a model to add or remove it

Ours

Partial · ANLS 0.85

Make ACOG poster and ship it to A.

34 characters12 tokens

Aux 2015

Wrong

Action # Description Proposed

Person Comments End Date Responsible Make ACOG poster and ship it to A. 4/26/04 DW In progress Conti 2 Identify the involvement of top opinion 4/16/04 DW

184 characters64 tokens

PiT-FT 2015

Wrong

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

120 characters64 tokens

ChronoGPT 2015

Wrong
Action # Description
Action # Description
Action # Description

The same 21-character fragment repeats 9 times until the token limit. Showing the first three.

209 characters39 tokens