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 · pagegfjv0228_3.png

What is the title of table 4?

OCR text of the page · 1,139 characters; the scanned image itself is not published
3
Betanal herbicides + HOE-23408 gave excellent postemergence
weed control at several locations particularly from applications
made by Wicks (North Platte Station, University of Nebraska).
Herbicide Effect on Production
Under normal production practices, herbicides have a beneficial
effect on crop production (Table 4) .
Table 4. Preplant herbicide effect on yields, Longmont, Colo., 1974
1b / A
Wt.
%
Herbicide
dose.
T/A Sugar
Purity
Nortron
3.5
26.0
17.3
92.4
Ro-Neet
4
24.9
17.1
92.1
Pyramin
25.0
17.2
91.5
Untreated
0
23.0
17.1
91.6
Similar results have been obtained previously.
Programmed Weeding Systems
Complete chemical weeding is the key to final stand planting and
elimination of clean up labor in beet fields. Programmed or
sequence herbicide applications offer the best chance for weed-
free beets and production improvement. It is estimated that only
: 15-20 percent of Great Western acreage was sequence treated in
.1974. A 40 percent goal achievement should be realized in 1975.
Consult the Sugarbeet Production Guide, page 29 for typical
sequence applications.
Source: https://www.industrydocuments.ucsf.edu/docs/gfjv0228
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
Preplant herbicide effect on yields, Longmont,Colo., 1974
Models
4 of 4 columns · click a model to add or remove it

Ours

Partial · ANLS 0.98

Preplant herbicide effect on yields, Longmont, Colo., 1974

58 characters19 tokens

Aux 2015

Wrong

Programmed Weeding Systems

26 characters6 tokens

PiT-FT 2015

Wrong

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

120 characters64 tokens

ChronoGPT 2015

Wrong

3.5 26.0 17.3 92.4 Ro-Neet 4.0 24.9 17.1 91.5 Untreated 0 23.0 17.1 91.6 Similar results have been obtained previously. Programmed

130 characters64 tokens