metadata

title: Open FinLLM Leaderboard
emoji: 🥇
colorFrom: green
colorTo: indigo
sdk: gradio
sdk_version: 4.42.0
app_file: app.py
pinned: true
license: apache-2.0

Open Financial LLM Leaderboard (OFLL)

The growing complexity of financial language models (LLMs) demands evaluations that go beyond general NLP benchmarks. Traditional leaderboards often focus on broader tasks like translation or summarization, but they fall short of addressing the specific needs of the finance industry. Financial tasks such as predicting stock movements, assessing credit risks, and extracting information from financial reports present unique challenges, requiring models with specialized capabilities. This is why we created the Open Financial LLM Leaderboard (OFLL).

Why OFLL?

OFLL provides a specialized evaluation framework tailored specifically to the financial sector. It fills a critical gap by offering a transparent, one-stop solution to assess model readiness for real-world financial applications. The leaderboard focuses on tasks that matter most to finance professionals—information extraction from financial documents, market sentiment analysis, and financial trend forecasting.

Key Differentiators

Comprehensive Financial Task Coverage: Unlike general LLM leaderboards that evaluate broad NLP capabilities, OFLL focuses exclusively on tasks directly relevant to finance. These include information extraction, sentiment analysis, credit risk scoring, and stock movement forecasting—tasks crucial for real-world financial decision-making.
Real-World Financial Relevance: OFLL uses datasets that represent real-world challenges in the finance industry. This ensures models are not only tested on general NLP tasks but are also evaluated on their ability to handle complex financial data, making them suitable for industry applications.
Focused Zero-Shot Evaluation: OFLL employs a zero-shot evaluation method, testing models on unseen financial tasks without prior fine-tuning. This highlights a model’s ability to generalize and perform well in financial contexts, such as predicting stock price movements or extracting entities from regulatory filings, without being explicitly trained on these tasks.

Key Features of OFLL

Diverse Task Categories: OFLL covers tasks across seven categories: Information Extraction (IE), Textual Analysis (TA), Question Answering (QA), Text Generation (TG), Risk Management (RM), Forecasting (FO), and Decision-Making (DM).
Robust Evaluation Metrics: Models are assessed using various metrics, including Accuracy, F1 Score, ROUGE Score, and Matthews Correlation Coefficient (MCC). These metrics provide a multidimensional view of model performance, helping users identify the strengths and weaknesses of each model.

The Open Financial LLM Leaderboard aims to set a new standard in evaluating the capabilities of language models in the financial domain, offering a specialized, real-world-focused benchmarking solution.

Contribute to OFLL

To make the leaderboard more accessible for external contributors, we offer clear guidelines for adding tasks, updating result files, and other maintenance activities.

Primary Files:
- src/env.py: Modify variables like repository paths for customization.
- src/about.py: Update task configurations here to add new datasets.
Adding New Tasks:
- Navigate to src/about.py and specify new tasks in the Tasks enum section.
- Each task requires details such as benchmark, metric, col_name, and category. For example:
```
taskX = Task("DatasetName", "MetricType", "ColumnName", category="Category")
```

Updating Results Files:

Results files should be in JSON format and structured as follows:

{
    "config": {
        "model_dtype": "torch.float16",
        "model_name": "path of the model on the hub: org/model",
        "model_sha": "revision on the hub"
    },
    "results": {
        "task_name": {
            "metric_name": score
        },
        "task_name2": {
            "metric_name": score
        }
    }
}

Updating Leaderboard Data:
- When a new task is added, ensure that the results JSON files reflect this update. This process will be automated in future releases.
- Access the current results at Hugging Face Datasets.
Useful Links:
- Hugging Face Leaderboard Documentation
- OFLL Demo on Hugging Face

If you encounter problem on the space, don't hesitate to restart it to remove the create eval-queue, eval-queue-bk, eval-results and eval-results-bk created folder.

Code logic for more complex edits

You'll find

the main table' columns names and properties in src/display/utils.py
the logic to read all results and request files, then convert them in dataframe lines, in src/leaderboard/read_evals.py, and src/populate.py
teh logic to allow or filter submissions in src/submission/submit.py and src/submission/check_validity.py

License

Distributed under the Apache License, Version 2.0.

SPDX-License-Identifier: Apache-2.0

Current submissions are manully evaluated. Will open a automatic evaluation pipeline in the future update

tags:

leaderboard
modality:text
submission:manual
test:public
judge:humans
eval:generation
language:English