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 · pagemymk0226_9.png

What is the 20th item in the list?

OCR text of the page · 1,002 characters; the scanned image itself is not published
VI
TOOLS OF IMPLEMENTATION
1.
Regular News Releases
2 .
Feature Story Development
3.
Mailings to Editorial writers
4 .
Mailings to Columnists
5.
TV News Clips
6 .
Radio Script Releases
7 .
Sugar and Health Press Kit
8 .
Literature Distribution
9 .
Media Seminars
10.
Media Convention Attendance
11.
Federal Government Liaison Representative
12.
State Government Liaison Representative
13.
Government Officials Seminars
14.
Legislative Surveillance Program
15.
Speakers Bureau
16.
Stock Speech with Slides
17.
Speech Forums/National and Regional
18.
Spokesman/Scientific
19.
Spokesman/Celebrity
20.
TV Talk Shows
21.
Film on Sugar and Health
22
Teachers Guide on Dental Health
23.
Generate and Publish Books on the Various Aspects of Sugar
24.
Letter-writing Campaigns
25 .
Sugar Information to Data Banks
26 .
Institutional Advertising (When necessary)
27,
Grant to Public Broadcasting Service
28.
Expand Association's Home Economics Program
Source: https://www.industrydocuments.ucsf.edu/docs/mymk0226
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
tv talk showsTV Talk shows
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

TV Talk Shows

13 characters4 tokens

Aux 2015

Wrong

26

2 characters3 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong

VI

2 characters1 tokens