Senior software engineer at Qualia Labs · Co-founder of Fox.Build Makerspace · Former co-founder of FarmBot

Asking llms to follow the i+1 rule of the input hypothesis

This is an LLM prompt I experimented with for language learning.

TIP

  • For good results, use a secondary prompt to perform translation.
  • I had good luck with temp/top_p/frequency penalties in the 0.6-0.8 range.

=== Example Input:
미국에서는 반대의 상황이 맞아요.

=== Examples Output:
반대의 상황이 맞아요.
미국에서는 반대의 주장이 맞아요.
미국에서는 같은 상황이 있어요.
한국에서는 반대의 상황이 맞아요.

=== Your Input: REPLACE THIS PART WITH AN EXAMPLE SENTENCE

=== Instructions

Use the given sentence to create new sentences by changing only one element at a time. The changes can include modifying, adding, or removing something from the sentence, but only one change is allowed. The sentence must also be natural and realistic.

Create variations of the original sentence by altering one element in each new sentence.
Ensure that the new sentences are grammatically correct and make logical sense.
Aim for diversity in the type of changes you make.

Create at least 10 new sentences following the rules outlined above, starting from the original sentence provided. Focus on variety and relevance in your modifications.


책상 위에 연필이 있어요. / There is a pencil on the desk.
책상 위에 필통 두 개 있어요. / There are two pencil cases on the desk.
책상 위에 필통이 없어요. / There is no pencil case on the desk.
책상 위에 필통이 있었어요. / There was a pencil case on the desk.
책상 위에 필통을 뒀어요. / I put a pencil case on the desk.
책상 아래에 필통이 있어요. / There is a pencil case under the desk.
책상 위에 필통이 있습니다. / There is a pencil case on the desk. (formal)
책상 위에 필통이 있나요? / Is there a pencil case on the desk?