PromptTensor Prompt Bank v1.0.1: Curated English Prompt Dataset for LLM Research

PromptTensor Prompt Bank v1 is a curated public dataset of English prompts designed for LLM and NLP research, prompt engineering, and benchmarking. The dataset contains 7,040 prompts after filtering and deduplication. Each prompt is labeled with structured metadata, including domain, subdomain, intent, difficulty, output style, and length target.This release is intended for research and applied workflows such as prompt analysis, instruction pattern discovery, classification, benchmarking, and dataset engineering. The dataset includes user prompts only and does not include model outputs. Public files are provided in machine-friendly formats, including JSONL, CSV, and Parquet.The dataset was created and curated by PromptTensor through a combination of automated filtering, taxonomy alignment, and manual review. A canonical dataset page with documentation, schema details, and related links is available on PromptTensor. DOI-backed releases and public mirrors also be available on external platforms.