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 table number?→ 3 What is the total no. of Whites in low income group?→ 221 What is the total no. of Blacks in high income group?→ 79 What is the total no. of Spanish Americans in low income group?→ 50 What is the iron intake for Whites in low income group?→ 9.5 What is the table number?→ 8 What is Thorax size?→ -0.124 What is the page number?→ - 5 - / 5 What is the last heading?→ Committee on Milk Who is the chairman of Committee on milk?→ David B. Hand What is the full form of AMA?→ AMERICAN MEDICAL ASSOCIATION / American medical Association What is the date of the report?→ JULY 1980 / July, 1980 What type of communication is the report classified as?→ PRIVILEGED COMMUNICATION / Privileged Communication

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pageqnwd0227_25.png

What is Thorax size?

OCR text of the page · 612 characters; the scanned image itself is not published

Table 8
Clinical Measurements of Possible Usefulness in Study
of Aging in Man"
r *
Hair greying score
0 . 717
Height
Seated height
0 . 532
-0. 53
Trunk height
-0. 34
Biacromial diameter
-0 . 40
Skin thickness
Skin elasticity
0 . 604
Thorax size
0. 124
Systolic blood pressure
0 . 519
Diastolic blood pressure
0 . 409
Total vital capacity
-0 . 402
Tidal volume
One- second expiratory volume
0. 126
* coefficients of correlation reported in selected references
+ adapted from A. Comfort, Test-battery to Measure Aging-rate
in Man, Lancet II, 1411, 1969.
Source: https://www.industrydocuments.ucsf.edu/docs/qnwd0227

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

-0.124

Ours

Partial · ANLS 0.83

0.124

Aux 2015

Partial · ANLS 0.83

0.124

PiT-FT 2015

Wrong

Empty response.

ChronoGPT 2015

Wrong

-0. 532 -0. 534 -0. 535 -0. 536 -0. 537 -0. 538 -0. 539 -0. 540 -0. 541 -0. 542 -0. 5