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 · pagenqnn0081_7.png

What was the pack spinner capacity?

OCR text of the page · 1,218 characters; the scanned image itself is not published
Permanent Counter Displays (Continued)
PCD Configurations
Time Savers, New Orleans - 1977
Tested three different PCD configurations with varying capacities. Objective
was to determine which generated the highest share for displayed brands.
1. A continuous counter display (original PCD) with add-on units. Capacity
was 101 packs.
2. A four sided rotating display (pack spinner) capacity was 118 packs.
3. A smaller display with capacity for 30 packs.
Each display was loaded with various styles of VANTAGE, MORE and/or NOW.
Results
Share of market for the displayed items was 3.6% under each of the
configurations tested
Total cigarette sales were not significantly different for each
configuration.
PCD Effectiveness
Bonded Oil, Dayton - 1977
Tested the effect of having PCD's present on total RJR sales. A single PCD
with a capacity of 104 packs was used in the test.
Results
Sales of brand styles loaded on the PCD were 5.8% higher than when the PCD
was not used.
Incremental sales of display brand styles (purchases by smokers who did not
claim the display brands as their usual brand) was 2.7% higher with a PCD
present.
Total cigarette category sales increased 0.3% with displays (not
significant) .
51338 6589
5
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
118 packs118
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

118 packs

9 characters5 tokens

Aux 2015

Partial · ANLS 0.78

101 packs

9 characters5 tokens

PiT-FT 2015

Wrong

<|<|<|<|<|<|<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

64 characters64 tokens

ChronoGPT 2015

Wrong

PCD Configuration:

Permanent Counter Displays (Continued) PCD Configuration: Tested three different PCD configurations with varying capacities. Objective was to determine which generated the highest share for displayed brands.

  1. A continuous counter display (original PCD) with add-on units.
296 characters64 tokens