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 · pagepgxg0224_2.png

How many Rooms are reserved at World Trade Centre for 5/15/00?

OCR text of the page · 958 characters; the scanned image itself is not published
GS List
Reynolds
Tobacco Holdings, Inc.
Goldman Sachs 5/16 Conference
Proposed Agenda
Depart W-S:
2:00pm on Monday, 5/15/00
RJR Presentation:
9:45 to 10:40 (allow 15 minutes for Q&A)
Breakout session:
10:40 to ~11:15 or 11:30
Lunch:
12:00 to ~1:20 with Capital Research
One on one:
1:30 to 2:00 with Bank of America
One on one:
2:00 to 2:30 with Oppenheimer Funds, Inc.
One on one:
2:30 to 3:00 with Fidelity
Depart NY:
4:30pm on Tuesday, 5/16/00
Issues to Discuss
Review Powerpoint Presentation
AJS: Business fundamentals and Eclipse
CAB: Litigation
Talking points per slide still in progress
Presenters change their own slides? Or Keith handle?
7 Rooms are reserved at World Trade Center for 5/15/00
(AJS, KJL, CAB, TJP, LJB, DMN, and Keith)
Checking into location for early dinner on 5/15 near the World Trade Center
Anyone from RJR to G. Bible lunch presentation?
52979 7882
Analyst Visits.xis
Source: https://www.industrydocuments.ucsf.edu/docs/pgxg0224
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
7
Models
4 of 4 columns · click a model to add or remove it

Ours

Wrong

7 Rooms are reserved at World Trade Center for 5/15/00

54 characters18 tokens

Aux 2015

Wrong

5/15/00

Rooms reserved at World Trade Center for 5/15/00

GS List Reynolds Tobacco Holdings, Inc. Goldman Sachs 5/16 Conference Proposed Agenda Depart W-S: 2:00pm on

166 characters64 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong

World Trade Center

Early Dinner

Early Lunch

Early Lunch

Early Lunch

Early Dinner

Early Lunch

Early Lunch

Early Lunch

Early Lunch

Early Lunch

Early Lunch

Early Lunch

228 characters64 tokens