---
language:
- en
task_categories:
- text-classification
dataset_info:
  features:
  - name: sent1
    dtype: string
  - name: sent2
    dtype: string
  - name: labels
    dtype: int64
  splits:
  - name: test
    num_bytes: 381014
    num_examples: 2048
  download_size: 184976
  dataset_size: 381014
configs:
- config_name: default
  data_files:
  - split: test
    path: data/test-*
tags:
- chemistry
- pubchem
- chemteb
license: cc-by-nc-sa-4.0
pretty_name: PubChem AI Paraphrased Sentence Pair Classification
size_categories:
- 1K<n<10K
---
# PubChem AI Paraphrased Sentence Pair Classification

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 ([ChEBI](https://www.ebi.ac.uk/chebi/) source).