Dataset Viewer
Auto-converted to Parquet Duplicate
assistant
stringlengths
11
81
prompt
stringlengths
2
15
⌀
When can we go to the pool?
license
Ruby likes bikes, but Cole likes cows.
omniscient
My mouse is white.
immense
I want to drink juice.
substantial
Jake's fish is long.
north
Joe's box is long.
dirty
We saw a hard sheep at the playground!
soft
Liz threw a ball, and Rosa caught a frog.
idiotic
There is a white pig in the mud.
previous
When we get to the beach, we will hop.
order
I like mice.
magic
Yesterday, Owen skipped to the barn.
appliance
I like kites.
depend
When we get to the party, we will swim.
exclusive
Maya hugs my friend.
nut
Here comes a curly mouse!
capable
Can you play like a cow?
left
Look at a quiet pig!
six
My brother has a fuzzy bug.
intelligent
Jen is cleaning a hat right now.
gentle
I can hop high.
direction
Where is my lion?
swanky
My doll is fast.
ambiguous
Dan washes a lion.
birds
Is that your box?
alcoholic
Cole has three mice.
expert
How do you make popcorn?
stocking
The cold chicken is jumping on the farm.
charge
My card is round.
string
When can we go to the table?
pretend
Let's go to the beach!
expect
First I take a squirrel, and then I share a car.
side
Jack, your doll is so cute!
mark
Do you like pie?
sand
Is that your puppy?
shrill
Max walks to the zoo.
curtain
My uncle has a happy spoon.
desire
Joe is smiling now.
snatch
Who has my turtle?
fire
Ann is running now.
lace
Look at the orange bells!
print
Nina bought a cow, and Liz got a monkey.
hulking
Ben closes the car.
monkey
When can we go to the fort?
wonder
Nina kicks a ball.
trucks
Rosa took a bird, and Tom pulled a kite.
lettuce
I want toys, and you want turtles.
nail
Here comes a new fish!
respect
The hungry bird skips in the yard.
handy
The pink bird walked in the yard, and the new bear swam in the pond.
tomatoes
When can we go to the store?
arrive
Owen, your monkey is so wet!
grin
The loud bear is walking on the hill.
women
What is June doing?
wash
Kate went to the door, but Dan walked to the kitchen.
special
Why is the goat hungry?
murky
What is Emma doing?
scorch
Nora sees a duck.
rustic
When can we go to the tent?
harass
Bob likes to swim at the park.
leather
Let's walk to the tent!
madly
First I see a toy, and then I share a ball.
adjoining
Lucy is skipping today.
red
Look at the slow bells!
wobble
Yesterday, Liz ran to the tent.
ducks
Ben, your flower is so black!
afford
Nina has three bugs.
earn
I like a pig.
wind
Yesterday, Ava hopped to the tent.
same
While the mouse was jumping in the house, the sheep walked to the garden.
gabby
The dirty puppy walked on the bed, and the white kitten clapped in the house.
plantation
Jack pulled a horse, and Nora pushed a rabbit.
vase
Mia, your shoe is so little!
collect
I want sheep, and you want socks.
quarter
Look at a nice rock!
elite
The clean snake is hopping in the grass.
change
The quiet mouse laughed in the house, and the gray rabbit waved in the grass.
modern
I like popcorn.
stranger
Lucy is washing a shell right now.
tangible
It is sunny today.
plug
Rosa washed a bird, and Tim drew a bird.
inexpensive
Bob has three sticks.
smoke
We saw a glad dog at the swing!
rot
While the worm was hopping in the grass, the turtle swam to the pond.
uptight
Do you like tigers?
various
Ella said, "Yay."
ludicrous
Do you like snakes?
copper
My book is big.
idea
Do not make a sandwich!
awful
Ava's rock is happy.
hurried
We had fun at the door!
preach
Jen went to the pond, but Dan walked to the table.
fanatical
Maya is crying now.
comb
Max swam to the pool, but Paul swam to the pool.
achiever
My kite is big.
snotty
The funny puppy sings on the bed.
tremendous
Where is my monkey?
afraid
Can you sleep like a chicken?
mountain
I want flowers, and you want books.
ignore
Look at a curly bug!
efficacious
End of preview. Expand in Data Studio

Random Sentences Dataset (version 2)

This is a random sentence dataset, it has random simplistic children sentences paired with completely unrelated random words. This dataset was used in SmolBabble2-360m, an AI model that spits out random sentences regardless of what is said to it. It is an improved version of the previous dataset, and it has much more examples. You can find more information on both of these models on the provided repository links.

The dataset has prompt and assistant pairs, and has ~9900 rows.

The script used to generate the sentences was mostly AI generated (because I do not know python), and it is in this repo. It can be expanded and is very fast.

Why?

This was made just for fun, but if you do find an actual use case for this dataset or just want to ask or say something, please open an issue and tell me!

Downloads last month
87