---
library_name: transformers
license: apache-2.0
base_model:
- Qwen/Qwen3-8B
tags:
- finance
model-index:
- name: ODA-Fin-SFT-8B
  results: []
datasets:
- OpenDataArena/ODA-Fin-SFT-318k
language:
- en
- zh
metrics:
- accuracy
- f1
pipeline_tag: question-answering
---

# Unlocking Data Value in Finance: A Study on Distillation and Difficulty-Aware Training



<div align="center">

[![Paper](https://img.shields.io/badge/arXiv-Paper-red)]()
[![Collections](https://img.shields.io/badge/🤗-Collections-yellow)]()

</div>


![Alt text](main_performance.png)


## 📖  Model Summary

ODA-Fin-SFT-8B is an 8B financial large language model fine-tuned on **ODA-Fin-SFT-318k**, a high-quality distilled financial dataset with strong Chain-of-Thought reasoning capabilities.

It is built on **Qwen3-8B** and optimized for financial understanding, sentiment analysis, and numerical reasoning over text and tables.


## 📈 Performance

ODA-Fin-SFT-8B achieves strong performance across 9 financial benchmarks.


![Alt text](main_results.png)

<!-- ### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 1
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 16
- total_train_batch_size: 128
- total_eval_batch_size: 64
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 3.0

### Framework versions

- Transformers 4.55.0
- Pytorch 2.6.0+cu124
- Datasets 3.2.0
- Tokenizers 0.21.0 -->