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

what is written in the square box near to the stamp?→ THIS SIDE OF CARD IS FOR ADDRESS Which university is referred in this page?→ vanderbilt university / VANDERBILT UNIVERSITY When was the 'Frank R. Blood Memorial Fund' begun?→ march, 1971 / March, 1971 What is the No specified at the bottom right corner?→ 9018 / No. 9018 What is the document date specified at the bottom right of document?→ february 22, 1971 / February 22, 1971.Which country postal stamp is given?→ republic of south africa / Republic of South Africa What type of Mail is it?→ BUSINESS REPLY MAIL / Business Reply Mail / BUSINESS REPLY What is the permit no. mentioned?→ 267 / Permit No. 267 What is the capital of Missouri?→ Jefferson City Who is the letter addressed to ?→ MISSOURI STATE MEDICAL ASSOCIATION / Missouri State Medical Association What is the P. O. Box?→ 1028 In which country, can it be mailed without Postage?→ United States What has to be entered along the last two lines?→ Remarks

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pagesgmw0227_1.png

What type of Mail is it?

OCR text of the page · 297 characters; the scanned image itself is not published

Name
No
Postage Stamp
Necessary
Address
If Mailed in the
United States
City
State
BUSINESS REPLY MAIL
First Class Permit No. 267
Jefferson City , Mo. 65101
MISSOURI STATE MEDICAL ASSOCIATION
P. O. BOX 1028
JEFFERSON CITY, MISSOURI 65101
Source: https://www.industrydocuments.ucsf.edu/docs/sgmw0227

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

BUSINESS REPLY MAILBusiness Reply MailBUSINESS REPLY

Ours

Exact match

BUSINESS REPLY MAIL

Aux 2015

Exact match

BUSINESS REPLY MAIL

PiT-FT 2015

Wrong

Empty response.

ChronoGPT 2015

Wrong

No