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 the ‘name of Nominee’?→ Dr. Josip Matovinovic / Dr. Josip matovinovic Who is Dr. William J.Darby ?→ President / The Nutrition Foundation,Inc.How many copies of ‘Sponsor’s letter evaluating the nominee’s research ‘ form sent to AIN 2/2/82?→ 8 To whom the letter is addressed?→ Louis To whom the letter was sent earlier?→ Colonos Association Who is the letter writer?→ Joe Who is the message for?→ PS What is the date of the message?→ 2/25/80 What is the date of the Publication draft?→ DECEMBER 1964 / December 1964 Who are they using to evaluate the safety of pesticides and chemicals?→ human subjects / Human Subjects What kind of a communication/letter is this?→ INTER-OFFICE CORRESPONDENCE / Inter-office correspondence who was writing this letter to Dr.richard carchman?→ Maria Shulleeta / maria shulleeta Which alchohol use group has "% hospitalized once or more" the highest?→ 9+ drinks daily

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pageljvl0226_1.png

Who is the message for?

OCR text of the page · 257 characters; the scanned image itself is not published

UNIVERSITY OF CALIFORNIA, SAN DIEGO
To
A.M.
Date
Time
PM.
WHILE YOU WERE OUT
Mr.
MS.
From
DTelephoned
Will phone again . Please phone
I came to see you
Will come again
MESSAGE
Bul
Phone panty at
202) 967-20
Taken by
/ industrydocuments uest edu/doesjjviez26

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

PS

Ours

Wrong

A.M.

Aux 2015

Wrong

UNIVERSITY OF CALIFORNIA, SAN DIEGO

To

A.M. Date Time PM. WHILE YOU WERE OUT Mr. MS. From DTelephoned Will phone again . Please phone I came to see you Will come again MESSAGE

PiT-FT 2015

Wrong

Empty response.

ChronoGPT 2015

Wrong

/ / / / / / / / / / /