This dataset consists of sentence pairs, with labels indicating their relationship: a positive label (1) means that one sentence is an AI-generated paraphrase of the other, while a negative label (0) indicates that the two sentences are distinct and unrelated. The paraphrases are created using the anthropic.claude-3-5-sonnet-20240620-v1:0 model on AWS Bedrock. The dataset is built from entity descriptions sourced from PubChem… See the full description on the dataset page:
https://huggingface.co/datasets/BASF-AI/PubChemAISentenceParaphrasePC.