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 · pagejrcy0227_94.png

What is the date of issue?

OCR text of the page · 953 characters; the scanned image itself is not published
I ACKNOWLEDGE RECEIPT OF TICKET(S) AND/
Page 7
Page 8
OR COUPONS FOR RELATED CHARGES DESCRIBED
UNIVERSAL CREDIT CARD CHARGE FORM
on reverse
DATE AND PLACE OF ISSUE
ON. PAYMENT IN FULL TO BE MADE WHEN
OR IN EXTENDED PAYMENTS IN ACCORD-
EASTERN iNES 007
CONTRACTOR INVOICE COPY
ANC
ITH STANDARD POLICY OF COMPANY ISSUING
IF EXTENDED PAYMENT DESIRED.
CARD AND AS REFLECTED IN APPLICABLE TARIFFS.
22 MAY 74
CIRCLE NO. OF MONTHS
EASTERN AIRLINES
X
DATE OF ISSUE
3
6
9
12
IN FLIGHT SERVCES
NAME OF PASSENGER IF OTHER THAN CARDHOLDER
OTATO NO.
COMPLETE ROUTING
CONNECTION OF PASSENGER WITH SUBSCRIBER
FARE BASIS
CARRIER
APPROVAL CODE
FROM
TICKETS NOT TRANSFERABLE
NO CASH REFUNDS
EITHER DIRECTION
AIRLINE
FORM
CREDIT CARD NAME/CODE
SERIAL NO. FORM OF CREDIT
To
BETWEEN
EA
NYC/NEWARK
007: 409 3: 568:325
UAQ 53 12 09
CITIZENS COMMISSION
WASH DC
DR CHANNING H LUSHBOUGH
EXP. DATE
ROUTE CODE
TAX
27.44
SC . 64
- 20
4093568325 0 0
FOU
2
29.64
850014
0165312090
AMY
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
22 MAY 74
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

22 MAY 74

9 characters7 tokens

Aux 2015

Wrong

22 MAY 74

CIRCLE NO. OF MONTHS

EASTERN AIRLINES

X

DATE OF ISSUE

3

6

9

12

IN FLIGHT SERVICES

NAME OF PASSENGER IF OTHER THAN CARDHOLDER

OTATO NO.

COMPLETE ROUTING

173 characters64 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong
- Date of issue
- Date of issue
- Date of issue

The same 16-character fragment repeats 12 times until the token limit. Showing the first three.

207 characters64 tokens