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.

7 of 5,330 prompts

what is the page no or page mentioned ?→ 2 / page 2 What is the name of the University?→ Vanderbilt university / VANDERBILT UNIVERSITY What is the Expenses for Salaries for 1986?→ $123,110 What is the Expenses for Publications for 1987?→ 10,596 / $10,596 What is the Expenses for Publications for 1986?→ 3,700 / $3,700 What is the Expenses for Salaries for 1987?→ $117,047 What is the Expenses for Employee Benefits for 1987?→ $18,559 / 18,559

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pageffbg0227_2.png

what is the page no or page mentioned ?

OCR text of the page · 860 characters; the scanned image itself is not published

Robert E. Shank, M.D.
Page 2
April 17, 1972
As you know, each candidate selects the area of special emphasis
to be covered in depth in his examinations. We are attempting to compile
for each area topics or rubrics within which questions may be developed.
We should like to express our appreciation for your response to Dr. Harold
Ellingson's request for help in this regard. The suggestions from last
fall's examiners have given us an excellent start in this undertaking.
We hope, too, to have a list of recommended essential readings for each
area, and will appreciate receiving your suggestions for such a list.
Please accept my personal thanks and those of the Board for your
cooperation.
Sincerely,
Bill
William P. Richardson, M. D.
Vice Chairman, General Preventive
Medicine
WPR :m jm
Enclosure
Source: https://www.industrydocuments.ucsf.edu/docs/ffbg0227

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

2page 2

Ours

Exact match

Page 2

Aux 2015

Exact match

PiT-FT 2015

Wrong

Empty response.

ChronoGPT 2015

Wrong

### General Preventive Medicine

### General Preventive Medicine

### General Preventive Medicine

The same 33-character fragment repeats 9 times until the token limit. Showing the first three.