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

What does NL denote?→ night letter / Night Letter Whose death is mentioned in the telegram?→ DR HEATH / Dr Heath Who should be notified about the death?→ BRANCH MANAGER / Branch manager What information will be wired later?→ funeral arrangements / Funeral arrangements What is the Total Volume?→ 661.1 What is the "Conversion Rate" for "Retail BSGSF"→ 50 What is the "Conversion Rate" for "Event"?→ 8 What is the "Volume (MM)" for "Retail BSGSF"?→ 172.8 What is an important factor in assuring accuracy ?→ THE DESIGN OF VOLUMETRIC APPARATUS / The design of volumetric apparatus What is the main title of the page?→ Food Chemicals Codex How many parts of water is present in fluid A?→ 4.4 How many parts of water is present in fluid T?→ 3.6 How many parts of ferric chloride is present in fluid A?→ 0.4

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pagexzmy0225_12.png

What is the "Conversion Rate" for "Event"?

OCR text of the page · 542 characters; the scanned image itself is not published

SALEM 2003 Relaunch Plan
52951 1267
Incremental Volume Rationalization
% Incremental/ % New Buyers/
Program
Reach (M)
Competitive
Conversion Rate
Buyers (M) Volume (MM)
Retail BSGSF
28,800
30
50
4,320
172.8
SL Tins
6,780
80
NA
5,500
109.9
Multibrand Retail
278
65
16
44
136.6
Direct Mail PPT
861
100
6
52
79.5
Event
1, 171
65
8
61
150.7
Total Volume
661.1
MARLIN Share Supported
.18
Volume upsides:
. Discounting expansion
GL promotion
Box conversion (competitive and franchise)
9
Source: https://www.industrydocuments ucsfedu/docs./vzmy0225.

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

Ours

Exact match

Aux 2015

Wrong

150.7

PiT-FT 2015

Wrong

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

ChronoGPT 2015

Wrong