README.md

---
license: cc-by-3.0
task_categories:
- text-generation
language:
- en
tags:
- agent
- finance
- multimodal
- spreadsheet
- workflow

configs:
  - config_name: Finch_Dataset_All
    data_files:
    - split: valid
      path:
        - Finch_Dataset_valid.jsonl
    - split: test
      path:
        - Finch_Dataset_test.jsonl
---

![Finch cover figure](figs/finch_workflow.jpeg)

# Finch: Benchmarking Finance & Accounting Workflows around Multimodal Enterprise Spreadsheets

This repository contains the dataset for **Finch**, an enterprise-level benchmark for evaluating an agent’s ability to act like a skilled finance & accounting expert on real-world workflows.

* **Paper**: _to be added_
* **Project Page**: https://huggingface.co/datasets/FinWorkBench/Finch  
* **Code**: https://github.com/FinWorkBench

---

## Dataset Description

Finch focuses on **composite finance & accounting workflows** that span:

> data entry/import, structuring/formatting, web search, cross-sheet/file retrieval, calculation, financial modeling, validation, translation, visualization, and reporting.

The workflows are derived from **real-world enterprise workspaces**, including:

- Large and messy **spreadsheets** with multimodal artifacts including text, tables, formulas, charts, pivots, images, etc 
- Linked **PDFs and documents** that provide additional business context  

We adopt a three-step workflow labeling process:

1. **Summarizing workflow types** supported by real collaborative enterprise email threads.  
2. **Deriving concrete workflow instances** from versioned spreadsheets and related files using LLMs.  
3. **Meticulous expert annotation** of instructions and reference outputs, involving hundreds of hours of expert work.

This process yields **172 enterprise-grade single- and multi-task workflows** with carefully written instructions and aligned input/reference files, capturing the intrinsic **complexity, messiness, and multimodality** of real-world finance & accounting work.

---

## Dataset Structure

The instruction-tuning corpus is released in **JSONL** format and follows the standard **LLaMA Factory Alpaca** schema.  
Each line is one **workflow-centric example**:

```json
{
  "instruction": "Task instruction for a specific finance & accounting workflow (e.g., 'Reconcile the cash balance and explain the variance using the provided spreadsheet tabs.').",
  "input": _to be added_,
  "output": _to be added_
}