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 · pagemtnh0227_5.png

What is the Permit No given in this document?

OCR text of the page · 1,476 characters; the scanned image itself is not published
POST OFFICE DEPARTMENT
PERMIT NO.
STATEMENT OF MAILING MATTER WITH PERMIT IMPRINTS
MAILER: Prepare in duplicate, completing all items by typewriter, pen or indelible pencil.
357
NUMBER OF
POST OFFICE
RECEIPT NO.
SACKS
CARTONS
OUTSIDE
BUNDLES
1st-Letters,
3rd-Circulars and
3rd-Books or cata-
written matter,
2nd-Newspapers
and magazines
other printed matter,
logs of 24 pages
4th-Parcel post, books,
merchandise, printed
post cards.
entered as
merchandise less
or more, seeds,
second-class mail.
than 16 ozs.
etc., less than 16 ozs.
matter, 16 ozs. or
more.
NAME AND ADDRESS OF PERMIT HOLDER
(Include ZIP Code)
TELEPHONE NO.
Check here when mailing is by authorized nonprofit organization
(Section 134.5 P.M.)
great western ian
YES
Mailer (other than authorized nonprofit organization) must
47NO
check here whether his total mailings made at bulk third-
class rates at all post offices, under any name, for the cur-
rent calendar year, exceed 250,000 pieces.
Weight of a single piece.
OZS.
Number of pieces in a pound. (For third-class
SIGNATURE OF PERMIT HOLDER OR AGENT
matter mailed at bulk rates. )
16
Bryant
Total number of PIECES in mailing.
15 62
NAME AND ADDRESS OF INDIVIDUAL OR ORGANIZATION FOR WHICH MAIL-
ING IS PREPARED (If other than permit bolder)
Total number of POUNDS in mailing.
97. 6
Rate chargeable on each (check one)
POUND
PIECE is
3.8
TOTAL COMPUTED POSTAGE $ 5 9. 36
POD Form
July 1969
3602
Source: https://www.industrydocuments.ucsf.edu/docs/mtnh0227
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
357
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

357

3 characters4 tokens

Aux 2015

Wrong

The Permit No. is 357.

26 characters11 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong
### Permit No given in this document

### Permit No given in this document

### Permit No given in this document

The same 38-character fragment repeats 6 times until the token limit. Showing the first three.

241 characters64 tokens