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 is the 'total respondents' in the Control cell?→ (299) / 299 What is the 'total respondents' in the Test cell?→ (276) / 276 What does the symbol * indicate?→ control vs. test significantly different at the 80% confidence level. / Control vs. Test significantly different at the 80% confidence level.From the 'total respondents' in the control cell, what % of them 'recall receiving the offer'?→ 81 What is the second base used for the statistical test?→ recall offer / Recall offer What is the first base used for the statistical test?→ total respondents / Total respondents What % of the test cell respondents, have 'correct message recall'?→ 11 What is the hand written number?→ 1040 What is the name of the association?→ THE INDIANA DIETETIC ASSOCIATION / The Indiana Dietetic association What is the department?→ Home Economics Department / Home Economics department What is zip code?→ 47809 What is the name of University?→ Indiana State University What is the name of the state?→ Indiana

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pagenynn0081_8.png

What % of the test cell respondents, have 'correct message recall'?

OCR text of the page · 547 characters; the scanned image itself is not published

WINSTON RIGHT DECISIONS-RIGHT NOW
Key Measures Summary
Control
Test
Significance
Cell
Cell
Testing
Base - Total Respondents
(299)
(276)
%
Ever Tried
94
97
Recent Trial
28
18
Ever Purchased
77
80
Recent Purchase
20
13
Recall Receiving Offer
81
61
Offer Changed Opinion of WINSTON
18
16
Base - Recall Offer
(241)
(167)
Used Offer
38
50
Positive Usage (Net)
65
63
Recall Insert (Net)
N/A
38
Correct Message Recall
N/A
1 1
Insert Changed Opinion of WINSTON
N/A
* Control vs. Test significantly different at the 80%% confidence level.
Page 6
51338 5943

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

N/A

PiT-FT 2015

Wrong

Empty response.

ChronoGPT 2015

Wrong

Control:

Test:

Correct Message Recall

(289) (276) % Ever Tried 94 97

Key Measures Summary

Control Test Significance Cell Base - Total Respondents (299) (176) (