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 · pagefgvh0227_91.png

What is the total number of participants lost to follow-up in Baltimore?

OCR text of the page · 625 characters; the scanned image itself is not published
Table 516
Number of Participants Currently Considered Lost
to Follow-up, by Clinical Center
Number of Participants Lost to Follow-up
Clinical Center
SI
UC
Total
A-Baltimore
B-New York
C-Minneapolis
D-San Francisco
E-St. Louis
F-Boston Harvard
G-Davis
H-Chicago-Northwestern
I-Chicago St. Joseph
J-University of Chicago
K-Portland
L-Boston University
M-Dayton
N-Newark
0-Philadelphia
P-Birmingham
Q-Pittsburgh
R-Chicago-Rush
S-Rutgers
T-Columbia
U-Los Angeles
V-Miami
WUOHHOOHOOOHOOHH HOHOWO
All Clinics Combined
19
IWAOHOOHOOHOH HOOHOOOOWO
U ODONHOHHOHONHOHNHOHOMC
Source: https://www.industrydocuments.ucsf.edu/docs/fgvh0227
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
Models
4 of 4 columns · click a model to add or remove it

Ours

Wrong

19

2 characters3 tokens

Aux 2015

Wrong

19

2 characters3 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong

Question:

What is the total number of participants lost to follow-up in Baltimore?

OCR text: Table 516 Number of Participants Currently Considered Lost to Follow-up

SI UC

A- B- C- D- E-

192 characters64 tokens