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 · pagekjmc0079_3.png

19.1 % of the responses fall under which category?

OCR text of the page · 176 characters; the scanned image itself is not published
Chemical Compounds Found In Cigarettes (First Response)
Don't Know 1.6
12.3
Extremely Concerned
Not At All Concerned 32.7
19.1
Very Concerned
34.3
Somewhat Concerned
51450 2641
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
very concernedVery concerned
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

Very Concerned

14 characters4 tokens

Aux 2015

Wrong

51450 2641

10 characters11 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong

Chemical Compounds Found In Cigarettes (First Response) Don't Know 1.6 12.3 Extremely Concerned 34.3 Somewhat Concerned 51450 2641

130 characters43 tokens