| 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). |