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 · pagentvw0217_12.png

When was ARHP- 41st Annual Meeting?

OCR text of the page · 1,180 characters; the scanned image itself is not published
Trimegestone - Publication and Abstract Tracking Report
7/30/03
III.
US Upcoming Meetings and Deadlines
CONTA213-015153
Date/
Deadline for
Meeting
Location
Scientific Abstract
Submission
NAMS (North American Menopause Society) -
September 18-20, 2003
May 1st, 2003
14th Annual Meeting
Miami Beach, FL
ASRM (American Society for Reproductive
October 11-16, 2003
May 1, 2003
Medicine) - 59th Annual Meeting
San Antonio, TX
SGI (Society for Gynecologic Investigation) -
March 24-27, 2004
October, 2003
51st Annual Scientific Meeting
Houston, TX
(Exact Date, TBD)
ACOG - 52nd Annual Clinical Meeting
May 1-5, 2004
September 26, 2003
Philadelphia, PA
(regular
October 17, 2003 (film
festival)
Endocrine Society - ENDO 2004
June 16-19, 2004
January, 2004
New Orleans, LA
(Exact Date, TBD)
ARHP - 41st Annual Meeting
September 7-12, 2004
June, 2004
Washington, DC
(Exact Date, TBD)
NAMS - 15th Annual Meeting
October 6-9, 2004
May, 2004
Washington, DC
(Exact Date, TBD)
ASRM - 60th Annual Meeting
October 16-21, 2004
TBD
Philadelphia, PA
Page [Page] of [Pages]
Design Write, Inc.
Source: https://www.industrydocuments.ucsf.edu/docs/ntvw0217
Confidential Pursuant to Confidentiality
Order
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
September 7-12, 2004
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

September 7-12, 2004

20 characters13 tokens

Aux 2015

Exact match

September 7-12, 2004

20 characters13 tokens

PiT-FT 2015

Wrong

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

120 characters64 tokens

ChronoGPT 2015

Wrong
### Confidentiality

### Confidentiality

### Confidentiality

The same 21-character fragment repeats 10 times until the token limit. Showing the first three.

226 characters64 tokens