README.md

metadata

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 source).