Finance-Instruct-500k is a financial reasoning dataset designed for training high-level language models for financial tasks, reasoning, and multi-turn dialogue.

The dataset contains more than 500,000 high-quality data in the financial field, covering financial question answering, reasoning, sentiment analysis, topic classification, multilingual named entity recognition and conversational AI.

Dataset features:

• Multi-round dialogue: Rich dialogue content, emphasizing contextual understanding and reasoning ability.
• Diverse data sources: Contains data from multiple high-quality datasets such as Cinder and Sujet-Finance-Instruct-177k.
• RAG format data: In the Retrieval Augmentation Generation (RAG) task, external data is appended before the user field to enhance context understanding.
• Deduplication and preprocessing: Eliminate overlapping and irregular entries to obtain cleaner, higher-quality data.
• XBRL Tagging: Contains structured financial entity tags from Financial-NER-NLP for advanced extraction tasks.