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 · pagehjcw0217_11.png

What is the page number?

OCR text of the page · 875 characters; the scanned image itself is not published
Medical and Scientific Communications Plan 2000: Premarin Family of Products
Pri-Med (Primary Medicine Today - West)
April 6-8, 2000
Long Beach, CA
AOA (American Osteopathic Association)
October 29-November 2, 2000
Orlando, FL
Allied Health
AANP (American Academy of Nurse Practitioners)
June 29-July 4, 2000
Washington, DC
AAPA (American Academy of Physician Assistants)
May 27-June 1, 2000
Chicago, IL
Specialists
NAMS (North American Menopause Society)
September 7-9, 2000
Orlando, FL
Follow-up suggestions at all meetings include:
1. Reprints related to the presented cases
2. Invitation to other programs
3. Representative follow-up visit
Medical Congress Activity Budget
Medical Congress Interactive Satellite Symposium (6 @ $200,000 each)
$1,200,000
11
DWRITE 066044
Jeffrey Solomon Exhibit SJosource: https://www.industrydocuments.ucsf.edu/docs/hjcw0217
Page 11 of 33
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
11
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

11

2 characters3 tokens

Aux 2015

Exact match

11

2 characters3 tokens

PiT-FT 2015

Wrong

<|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|>

120 characters64 tokens

ChronoGPT 2015

Wrong

OCTOBER 29, 2000 OCTOBER 29, 2000 OCTOBER 29, 2000 OCTOBER 29, 2000 OCTOBER 29, 2000 OCTOBER 29, 2000 OCTOBER 29, 2000 OCTOBER 29, 2000

135 characters64 tokens