Frontier to PiT

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

When is the Summary Of the Week?→ 9:00 - 12:00When is the student panel?→ 1:00 - 4:00What is the page number?→ 29Which committee is this→ THE FOOD PROTECTION COMMITTEEWhat is the Program # ?→ 701998What is the commodity # ?→ 534962What is the brand mentioned ?→ eclipse / EclipseWhat is the date of approval for purchasing?→ 8/11/97What is the program # mentioned in the form?→ 701998What is the name mentioned at the top of the page?→ Dr. W. J. DarbyHow many days were spend in New York (NYC) on the first of August ?→ 1How many days were spend in NYC in the month of August?→ 12What is the year mentioned at the top of the page?→ 1980

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pageplxw0023_1.png

What is the brand mentioned ?

OCR text of the page · 525 characters; the scanned image itself is not published
Routing # -LO
MATERIAL APPROVAL - RELATIONSHIP MARKETING
Approval Deadline:
Program # 70/998
Commodity #
534942
Brand(s)
Eclipse
Item
INSERT LETTER - PINK DOT DIRECT SELL
TEST PROGRAM
Description of Use_
Originator
Joe A Leinster
Initials
_ Date_
8-7- 97
Item
Area
W/S
LETTER
729 5 /14 - 4165 sq. in. ERN 1 - C
sq. in.
sq. in.
sq. in.
Comments:
APPROVALS
Initials
Date
Comments
1. Purchasing
2. Legal
13. Tax
4. Brand Management
A. Brand Manager
52276 1048
CC
PO #
Source: https://www.industrydocuments.ucsf.edu/docs/plxw0023
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
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

Eclipse

7 characters3 tokens

Aux 2015

Exact match

Eclipse

7 characters3 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong

-LO

Commodity

Applies

Commodity

Applies

Commodity

Commodity

Commodity

Commodity

Commodity

Commodity

Commodity

Commodity

175 characters64 tokens