đ xFinder-qwen1505
xFinder-qwen1505 is a model tailored for extracting key answers from large language models (LLMs). It enhances the evaluation of LLMs by providing accurate key answer extraction, addressing the limitations of traditional methods.
đ Quick Start
Although the original README doesn't provide a quick - start section, you can refer to the repository for more information on how to get started with xFinder-qwen1505.
⨠Features
- Specifically designed for key answer extraction in LLMs.
- Fine - tuned from Qwen - 1.5 - 0.5B to improve performance.
- Addresses the limitations of traditional RegEx - based extraction methods.
- Improves the reliability of model assessments across various tasks.
đĻ Installation
The original README does not contain installation steps, so this section is skipped.
đģ Usage Examples
The original README does not provide code examples, so this section is skipped.
đ Documentation
Model Details
xFinder-qwen1505 is a model specifically designed for key answer extraction in large language models (LLMs). It is trained by fine - tuning Qwen-1.5-0.5B.
Model Sources
- Repository: https://github.com/IAAR-Shanghai/xFinder
- Paper: https://openreview.net/forum?id=7UqQJUKaLM
Uses
xFinder is primarily used to enhance the evaluation of LLMs by accurately extracting key answers from their outputs. It addresses the limitations of traditional regular expression (RegEx) - based extraction methods, which often fail to handle the diverse and complex outputs generated by LLMs. xFinder improves the reliability of model assessments across various tasks.
Training Details
xFinder-qwen1505 is fine - tuned from Qwen-1.5-0.5B. The training data consists of approximately 26.9K samples from the Key Answer Finder (KAF) dataset. This dataset is designed to enhance the accuracy and robustness of key answer extraction and includes a variety of tasks. It has been meticulously annotated by GPT - 4 and human experts to ensure high - quality training and evaluation. For more details, see this paper and try it with code.
Evaluation
xFinder is evaluated on the fully human - annotated test and generalization sets of the KAF dataset. The results demonstrate significant improvements in extraction accuracy and robustness compared to traditional methods. For more details, please refer to the paper and try it out using the provided code.
đ§ Technical Details
The original README does not have detailed technical implementation information, so this section is skipped.
đ License
The model is released under the [cc - by - nc - nd - 4.0](https://creativecommons.org/licenses/by - nc - nd/4.0/) license.
đ Citation
@inproceedings{
xFinder,
title={xFinder: Large Language Models as Automated Evaluators for Reliable Evaluation},
author={Qingchen Yu and Zifan Zheng and Shichao Song and Zhiyu li and Feiyu Xiong and Bo Tang and Ding Chen},
booktitle={The Thirteenth International Conference on Learning Representations},
year={2025},
url={https://openreview.net/forum?id=7UqQJUKaLM}
}