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 · pagehsfn0020_2.png

Which is the first brand style surveyed here?

OCR text of the page · 745 characters; the scanned image itself is not published
of. 1900
FIELD SALES SURVEY MAY PROMOTIONS
WINSTON B2GIF
. Did retailers place the "No Bull 5" ballots box shipped in the promotional sku?
About 50% were placed by retailer
Did the Winston sleeve header card stand up when sleeves were placed in the display?
Yes
Did the retailers place the easel card? If not why?
SOME DID , COMMENTS - FORGOT ABOUT IT !
SALEM BIGIF
What % of the time was the Salem easel card placed?
25%
.
Where was the easel card placed? (e.g. next to the cash register)
On front counter or selling area.
What % of the time was the display placed?
75%
.
What % of the time were both the easel and the display placed?
35%
Was the ORB. e paster placed? 50%
52422 7221
Source: https://www.industrydocuments.ucsf.edu/docs/hsfn0020
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
WINSTON B2G1F
Models
4 of 4 columns · click a model to add or remove it

Ours

Partial · ANLS 0.92

WINSTON B2GIF

13 characters8 tokens

Aux 2015

Wrong

52422 7221

10 characters11 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Wrong
### Field Sales Survey

### Field Sales Survey

### Field Sales Survey

The same 24-character fragment repeats 10 times until the token limit. Showing the first three.

262 characters64 tokens