73 lines
2.7 KiB
Markdown
73 lines
2.7 KiB
Markdown
|
---
|
|||
|
|
language:
|
|||
|
|
- en
|
|||
|
|
license: llama3
|
|||
|
|
tags:
|
|||
|
|
- Llama-3
|
|||
|
|
- Financial Analysis
|
|||
|
|
- RL
|
|||
|
|
- Atropos
|
|||
|
|
- Fundamentals Prediction
|
|||
|
|
- Nous Research
|
|||
|
|
- reasoning
|
|||
|
|
- transformers
|
|||
|
|
- reinforcement learning
|
|||
|
|
- json mode
|
|||
|
|
base_model: meta-llama/Meta-Llama-3.1-8B
|
|||
|
|
library_name: transformers
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
# DeepHermes Financial Fundamentals Prediction Specialist - Atropos RL
|
|||
|
|
|
|||
|
|
## Model Overview
|
|||
|
|
|
|||
|
|
The **DeepHermes Financial Fundamentals Prediction Specialist - Atropos RL** is an experimental model artifact, fine-tuned by Nous Research using our new open source LLM RL Gym, Atropos. This model specifically aims to enhance the accuracy of financial fundamentals predictions through reasoning-intensive reinforcement learning techniques.
|
|||
|
|
|
|||
|
|
**Note**: This model is experimental and is not intended as a general-purpose state-of-the-art solution.
|
|||
|
|
|
|||
|
|
## Atropos Open Source Framework
|
|||
|
|
|
|||
|
|
Atropos is Nous Research's open-source reinforcement learning environment stack, engineered to optimize diverse LLM capabilities through structured RL methodologies. Contributions and active engagement from the community are highly encouraged:
|
|||
|
|
|
|||
|
|
🔗 [Atropos GitHub Repository](https://github.com/NousResearch/Atropos)
|
|||
|
|
|
|||
|
|
## Evaluation Results
|
|||
|
|
|
|||
|
|
Training and evaluation focused on improving financial fundamentals direction prediction accuracy:
|
|||
|
|
|
|||
|
|
| Evaluation Metric | Final Accuracy |
|
|||
|
|
| ------------------------------------- | ----------------------- |
|
|||
|
|
| Direction Prediction Accuracy (train) | \~20% -> \~50% Accuracy |
|
|||
|
|
|
|||
|
|
|
|||
|
|
|
|||
|
|
## Key Features
|
|||
|
|
|
|||
|
|
- **Financial Fundamentals Prediction**: Enhanced capability to predict market direction using RL during intensive reasoning tasks.
|
|||
|
|
- **Built with Atropos RL Environments**: Built with the open-source Atropos reinforcement learning stack.
|
|||
|
|
- **Reasoning-Enhanced Predictions**: Specifically optimized for scenarios involving deep analytical reasoning in financial contexts.
|
|||
|
|
|
|||
|
|
## Usage
|
|||
|
|
|
|||
|
|
The model is optimized for reasoning-intensive financial analysis tasks and supports:
|
|||
|
|
|
|||
|
|
- **Deep Reasoning Mode given a company's context (previous quarter's financials data) to predict the next future quarter's fundamental metric direction)**
|
|||
|
|
|
|||
|
|
## Community and Support
|
|||
|
|
|
|||
|
|
We welcome contributions, suggestions, and issues through our community channels:
|
|||
|
|
|
|||
|
|
- [Atropos Issues](https://github.com/NousResearch/Atropos/issues)
|
|||
|
|
|
|||
|
|
## How to Cite
|
|||
|
|
|
|||
|
|
```bibtex
|
|||
|
|
@misc{
|
|||
|
|
title={DeepHermes Financial Fundamentals Prediction Specialist - Atropos RL},
|
|||
|
|
author={Teknium and Dakota Mahan and Roger Jin and Chen Guang and Jai Suphavadeeprasit and Jeffrey Quesnelle},
|
|||
|
|
year={2025},
|
|||
|
|
url={https://huggingface.co/NousResearch/DeepHermes-Financial-Fundamentals-Prediction-Specialist-Atropos-RL}
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|