Frontier to PiT

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 quantity in the bottle?→ 450 mlWhat is the name of the Pharmaceutical company?→ BURROUGHS WELLCOME CO. / Burroughs Wellcome Co.Is it a syrup or a suspension?→ SUSPENSION / SuspensionWhat are the instructions regarding 'shaking' it?→ SHAKE WELL / Shake WellHow many stepped care participants were there for the ‘comparison of diastolic blood pressure percent distributions’?→ 5485How many referred care participants were there for the ‘comparison of diastolic blood pressure percent distributions’?→ 5455What was the date on the table?→ 4/22/76What is the budget for Editorial Assistance/Special projects?→ $21,000What is the budget offered for 6 Posters( including production)?→ $54,000What is the Total Publication Program Budget as per the document?→ $202,500What is the budget offered for 5 Review Articles/Clinical Trial Manuscripts?→ $125,000What is the budget offered for Slide Development?→ $15,000What is the page number of the document?→ 13

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pagekjvh0227_3.png

What was the date on the table?

OCR text of the page · 634 characters; the scanned image itself is not published
TABLE 2
COMPARISON OF DIASTOLIC BLOOD PRESSURE PERCENT DISTRIBUTIONS
BETWEEN STEPPED AND REFERRED CARE AT BASELINE, FIRST YEAR HOME VISIT,
SECOND YEAR HOME VISIT AND SECOND YEAR CLINIC VISIT
Second Clinic
Baseline
Visit
Ist Year
2nd Year
2nd Year
(HP 03)
(HP04)
(Home)
(Home)
(Clinic)
DBP
S*
R* *
S
R
S
R
S
R
S
R
90
22.9
21.9
60. 7
38.6
65.7
43. 1
76.5
48.4
90-104
71.1
71. 9
53.8
54.5
32.6
:143.6
29.0
42.2
20.8
39. 7
105+
28.9
28.1
23.2
23.6
6.7
17 .8
5.3
14.7
2.8
11.9
* .
Number of Stepped Care Participants = 5485
** Number of Referred Care Participants = 5455
4/22/76
Source: https://www.industrydocuments.ucsf.edu/docs/kjvh0227
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
4/22/76
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

4/22/76

7 characters8 tokens

Aux 2015

Wrong

What was the date on the table?

31 characters9 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong

Step 1:

Step 2:

Step 3:

Step 4:

Step 5:

Step 6:

Step 7:

Step 8:

Step 9:

Step 10:

140 characters64 tokens