RUC-NLPIR/OmniEval-HallucinationEvaluator

Dataset Information

We introduce an omnidirectional and automatic RAG benchmark, OmniEval: An Omnidirectional and Automatic RAG Evaluation Benchmark in Financial Domain, in the financial domain. Our benchmark is characterized by its multi-dimensional evaluation framework, including:

1. a matrix-based RAG scenario evaluation system that categorizes queries into five task classes and 16 financial topics, leading to a structured assessment of diverse query scenarios;
2. a multi-dimensional evaluation data generation approach, which combines GPT-4-based automatic generation and human annotation, achieving an 87.47% acceptance ratio in human evaluations on generated instances;
3. a multi-stage evaluation system that evaluates both retrieval and generation performance, result in a comprehensive evaluation on the RAG pipeline;
4. robust evaluation metrics derived from rule-based and LLM-based ones, enhancing the reliability of assessments through manual annotations and supervised fine-tuning of an LLM evaluator.

Useful Links: 📝 Paper • 🤗 Hugging Face • 🧩 Github

We have trained two models from Qwen2.5-7B by the lora strategy and human-annotation labels to implement model-based evaluation.Note that the evaluator of hallucination is different from other four.

We provide the evaluator for the hallucination metric in this repo.

🌟 Citation

@misc{wang2024omnievalomnidirectionalautomaticrag,
      title={OmniEval: An Omnidirectional and Automatic RAG Evaluation Benchmark in Financial Domain}, 
      author={Shuting Wang and Jiejun Tan and Zhicheng Dou and Ji-Rong Wen},
      year={2024},
      eprint={2412.13018},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2412.13018}, 
}