TheFinAI/FinTagging_BIO

FinTagging: An LLM-ready Benchmark for Extracting and Structuring Financial Information

FinTagging is the first full-scope, table-aware XBRL benchmark designed to evaluate the structured information extraction and semantic alignment capabilities of large language models (LLMs) in the context of XBRL-based financial reporting. It decomposes the XBRL tagging problem into two subtasks:

• FinNI: Financial entity extraction.
• FinCL: Taxonomy-driven concept alignment.

FinTagging requires models to jointly extract facts from both unstructured text and structured tables and align them with the full 10k+ US-GAAP taxonomy.

Paper | Evaluation Framework

This repository contains the original benchmark dataset without preprocessing. Annotated data (benchmark_ground_truth_pipeline.json) is provided in the "annotation" folder. For preprocessed datasets suitable for specific model architectures, please see the linked datasets in the Github README.

Datasets:

• FinNI-eval: Evaluation set for FinNI subtask.
• FinCL-eval: Evaluation set for FinCL subtask.
• FinTagging_BIO: BIO-format dataset for token-level tagging.