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 test-retest for BG1?→ .67 What is the test-retest for cognitive composite?→ .87 What is the test-retest for Sentence span?→ .60 What is the Interobserver for BG1?→ .97 What is the Interobserver for recognition?→ .99 What does 1 stand for?→ Lowest pentile What does 2 stand for?→ Middle 60% of scores What does 3 stand for?→ Highest pentile How much was the unexpended income at the end of first fiscal year?→ $315,135.46 How much was the unexpended balance at the end of second fiscal year?→ $268,007.47 How much was the total research expenses for second fiscal year?→ 140,167.30 How much amount was expended for 'Salaries' for second fiscal year?→ 52,169.83 How much amount was expended for 'Taxes' for second fiscal year?→ 867.33

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pageyyhd0227_61.png

What does 2 stand for?

OCR text of the page · 823 characters; the scanned image itself is not published

TABLE 4-B (CONT'D.)
RELATIVE RISK OF FALLING INTO EXTREME 20% ACCORDING TO SUPPLEMENTATION INGESTIC " .
AND SOCIOECONOMIC STATUS
VARIABLE : Composite Infant Scale 24 Month Motor Score
Sample Size (N's)
Percentages
Chi-Square
Test Performance Group
TOTAL SAMPLE
Low Med. 2 High3 Tot.
LOW
Med.
High
Tot.
d.f. e
Supplementation
65
127
45
237
27
54
19
100
20.4
4
L.01
34
131
53
218
16
60
24
100
Category :
N
6
49
25
80
8
61
31
100
105
307 123
535
20
57 .
23
LOW SES
34
55
20
109
31
50
18
100
11.3
4
2.05
63
32
113
16
56
28
100
N
5
26
40
13
65
23
100
and
57
144
61
262
22
55
23
HIGH SES
. 30
71
24
125
24
57
19
100
15.7
4
2.01
16
68
20
104
15
65
19
100
2
23
16
40
3
58
40
100
-and
47
162
60
269
17
60
22
1 = Lowest pentile
2 = Middle 60% of scores
3 = Highest pentile
Source: https://www.industrydocuments.ucsf.edu/docs/yyhd0227

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

Middle 60% of scores

Ours

Exact match

Middle 60% of scores

Aux 2015

Wrong

We are given a table that lists the **Relative Risk of falling into extreme 20% according to supplementation ingestic " and **Socioeconomic Status (SES) variable Composite Infant Scale 24 Month Motor Score.

We are to answer: "What does 2 stand for?"

Looking

PiT-FT 2015

Wrong

Empty response.

ChronoGPT 2015

Wrong

Table 4-B (CONT'D.) Relative Risk of Falling into Extreme 20% Accord

ChronoGPT

SES

Med.

Tot.

L.01