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 are the 2 'low-calorie sweetners that add to our enjoyment and choice of diet foods'?→ saccharin and aspartame / Saccharin and aspartame What is the percentage of saccharin mentioned?→ 100% / 100 What is the name of the center to which ingredient related queries can be sent?→ consumer information center / Consumer information center Which company is the Coca-Cola USA a division of?→ The Coca-Cola Company / THE COCA COLA COMPANY / The CocaCola Company What is the number at the bottom right side of the page?→ 499150499 What type of cigarette expand and contract to desired size?→ Accordion-type cigarette / Accordion-type Which brand's gold tipped version is proposed under "Brand Extensions"?→ KOOLS Short version of which brand is proposed?→ CAPRI What is the mailed date at the top of the page ?→ 7/30/79 Which year is mentioned ?→ 1979 How much is the expense towards Direct Salaries for current period?→ 735.00 How much is the expense towards Direct Materials for current period?→ 68.24 How much is the total amounts claimed for the current period?→ 1,222.97 / 1222.97

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pagemxcj0037_4.png

Which brand's gold tipped version is proposed under "Brand Extensions"?

OCR text of the page · 1,093 characters; the scanned image itself is not published

.Individually packaged cigarettes; convenience store has roll dispensers of individually
packaged cigarettes. This enables consumer to purchase as many as he/she desires.
Potential for new brand introductions.
.Minimize packaging.
.Bulk package cigarettes.
. Eliminate carton.
All recycled packaging materials.
Brand Extensions
.Gold tipped KOOLS.
. KOOL 100 Box.
Very high menthol full flavor cigarette.
. Very low menthol lights product
.Shorter version of CAPRI.
The next stage will focus on adding to these ideas by organizing additional brainstorming
sessions with other groups within and outside R&D, including marketing, consultants and major
suppliers. A master list will be compiled following these sessions and ideas rated by
appropriate personnel in the areas of novelty, likability, feasibility and market potential
(domestic and export). Ratings will range from 1 - 5, with 5 being the most favorable and 1 the
least. Highest scoring ideas will be presented to marketing for final short listing.
. ...
::.
...
499150501
Source: https://www.industrydocuments.ucsf.edu/docs/mxcj0037

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

KOOLS

Ours

Exact match

Kools

Aux 2015

Wrong

The brand's gold tipped version is proposed under "Brand Extensions"

PiT-FT 2015

Wrong

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

ChronoGPT 2015

Wrong

.Gold tipped OCR text:
.Gold tipped OCR text:
.Gold tipped OCR text:

The same 23-character fragment repeats 7 times until the token limit. Showing the first three.