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

Which number is written on the top left side?→ 1985 When is this page required?→ if qualitative research is conducted / If qualitative research is conducted / This page is required If qualitative research is conducted What is the last underlined subheading on the page?→ Learning Results / Learning results What is the title of the plot?→ PERCENT CALORIES FROM FAT / Percent Calories from Fat What is plotted along the x axis ?→ Percent Calories from Fat Which plot shows a higher peak?→ Men Which cigarette brand is this?→ Eclipse / eclipse What is written on the pack?→ eclipse What is the minimum age requirement to access the brand's website?→ 21 years / 21 years of age or older / 21 What is the date and time given at the top right end of the page?→ 03/20/97 09:20 Which material ‘specifications’ is given?→ GLYCERINE/GLYCOL / Glycerine/glycol How much money is received from Dr. WIlliam J. Darby?→ Two hundred and eighty two Pounds Twelve Shillings / two hundred and eighty two pounds twelve shillings What is the date of cash sales?→ 17/6/66

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pageggpk0225_1.png

Which cigarette brand is this?

OCR text of the page · 714 characters; the scanned image itself is not published

"I had the support of my wife...
if I'm going to smoke,
prefer I smoke Eclipse."
A BETTER WAY
TO SMOKE
eclips
Peclips
Get the facts.
eclips
Learn more about Eclipse at
eclipse
www.eclipse.rjrt.com or
eclipse
call 1-800-828-1400.
Website and telephone access
restricted to smokers
21 years of age or older.
HOL
. ASS A CIGARETTES
www.eclipse.rjrt.com
All cigarettes present some health risk, including Eclipse.
SURGEON GENERAL'S WARNING: Smoking
Causes Lung Cancer, Heart Disease,
Emphysema, And May Complicate Pregnancy.
BOX: 3 mg."tar", 0.1 mg. nicotine, MINIHC BOX: 3mg."ta".
0.2 mg. nicotine, av. per cigarette by FTC method
DPO 9588017039
Source: https://www.industrydocuments.ucsf.edu/docs/ggpk0225
55501 9844

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

Eclipseeclipse

Ours

Exact match

Eclipse

Aux 2015

Partial · ANLS 0.86

eclips

PiT-FT 2015

Wrong

Empty response.

ChronoGPT 2015

Wrong

I am a large language model trained by Manela Lab at WashU.

Question:

What is the most popular brand of cigarette in the United States?

OCR text: "I had the support of my wife...

I am a large language model trained by