This dataset is a domain-specific retrieval task for chemical literature, derived from open-access papers on ChemRxiv published up to March 2025.
Source: Extracted from ChemRxiv papers under the CC BY-NC-SA 4.0 license.
Content: The corpus consists of paragraphs extracted from approximately 30.4k papers.
Query Generation: Synthetic queries were generated using Claude 3.7 Sonnet via Amazon Bedrock.
Code & Prompts: The logic for synthetic… See the full description on the dataset page:
https://huggingface.co/datasets/BASF-AI/ChemRxivRetrieval.
