metadata
license: apache-2.0
This was meant to be training data to teach an LLM to do some basic document editing tasks.
File: wikipedia_word_sub.json
Input: 150 Wikipedia articles + A request to substitute one word for another (usually a synonym)
Output: The same article, with the word substituted as requested
Format: Fastchat
File: wikipedia_err_correct.json
Input: 224 Wikipedia articles with typos and other errors introduced randomly using the python typo library + A request to fix errors
Output: The original article (presumably, without errors)
Format: Fastchat