dangermouse77's picture
Librarian Bot: Add language metadata for dataset (#2)
9586689 verified
metadata
language:
  - en
license: cc-by-sa-4.0

It is the [SQuAD] & [TriviaQA (rc.wikipedia subset)] datasets concatenated but then inverted with answer first and then question.

Useful to train LLM models that given an answer need to generate a question. These AQ-models, AQ for AnswerQuestion are useful in conversations with other normal LLM bots that given a question generate an answer.

If you have an automatic conversation between two LLMs, one QA-LLM and one AQ-LLM the conversation will not get stuck but continue forever.

Last but not least: this dataset was created with help of python scripts suggested by ChatGPT-4o (using vibe programming as Karpathy names it)