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

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pagelnmm0081_7.png

What does the image of the skull represent?

OCR text of the page · 262 characters; the scanned image itself is not published
THESE
Alcoholics' .
ARE
ALL
Industrial Polluters
OK !
Drug Abusers
SMOKING
IS
Obesity
EVIL
AND
Criminals
TAXABLE . . .
AIDS
According To Hillary . .
SO YOU GET
51359 5454
Smokers Are Really
HIGHER TAXES
Bad And Evil People ! !
FOR SMOKING ! !
BULL PUCKEY #?!*@?!
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
aidsAIDS
Models
4 of 4 columns · click a model to add or remove it

Ours

Wrong

Bad and Evil People

19 characters5 tokens

Aux 2015

Wrong

The image of the skull represents the concept of "evil" and "taxation" in the context of the 1980s anti-smoking movement.

121 characters34 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong
Smoking is a bad habit
Smoking is a bad habit
Smoking is a bad habit

The same 23-character fragment repeats 9 times until the token limit. Showing the first three.

209 characters64 tokens