%0 Generic %1 yue2023automatic %A Yue, Xiang %A Wang, Boshi %A Chen, Ziru %A Zhang, Kai %A Su, Yu %A Sun, Huan %D 2023 %K attribution fact-checking hallucinations llms metrics %R https://doi.org/10.48550/arXiv.2305.06311 %T Automatic Evaluation of Attribution by Large Language Models %U https://arxiv.org/pdf/2305.06311 %X A recent focus of large language model (LLM) development, as exemplified by generative search engines, is to incorporate external references to generate and support its claims. However, evaluating the attribution, i.e., verifying whether the generated statement is fully supported by the cited reference, remains an open problem. Although human evaluation is common practice, it is costly and time-consuming. In this paper, we investigate the automatic evaluation of attribution given by LLMs. We begin by defining different types of attribution errors, and then explore two approaches for automatic evaluation: prompting LLMs and fine-tuning smaller LMs. The fine-tuning data is repurposed from related tasks such as question answering, fact-checking, natural language inference, and summarization. We manually curate a set of test examples covering 12 domains from a generative search engine, New Bing. Our results on this curated test set and simulated examples from existing benchmarks highlight both promising signals and challenges. We hope our problem formulation, testbeds, and findings will help lay the foundation for future studies on this important problem.

@misc{yue2023automatic, abstract = {A recent focus of large language model (LLM) development, as exemplified by generative search engines, is to incorporate external references to generate and support its claims. However, evaluating the attribution, i.e., verifying whether the generated statement is fully supported by the cited reference, remains an open problem. Although human evaluation is common practice, it is costly and time-consuming. In this paper, we investigate the automatic evaluation of attribution given by LLMs. We begin by defining different types of attribution errors, and then explore two approaches for automatic evaluation: prompting LLMs and fine-tuning smaller LMs. The fine-tuning data is repurposed from related tasks such as question answering, fact-checking, natural language inference, and summarization. We manually curate a set of test examples covering 12 domains from a generative search engine, New Bing. Our results on this curated test set and simulated examples from existing benchmarks highlight both promising signals and challenges. We hope our problem formulation, testbeds, and findings will help lay the foundation for future studies on this important problem. }, added-at = {2024-05-07T16:55:23.000+0200}, archiveprefix = {arXiv}, author = {Yue, Xiang and Wang, Boshi and Chen, Ziru and Zhang, Kai and Su, Yu and Sun, Huan}, biburl = {https://www.bibsonomy.org/bibtex/24c5ac5e73f9deb811cfee13661b9c204/ghagerer}, doi = {https://doi.org/10.48550/arXiv.2305.06311}, eprint = {2305.06311}, interhash = {3340c36378136074874ea7f21ec424b7}, intrahash = {4c5ac5e73f9deb811cfee13661b9c204}, keywords = {attribution fact-checking hallucinations llms metrics}, primaryclass = {cs.CL}, timestamp = {2024-05-07T16:55:23.000+0200}, title = {Automatic Evaluation of Attribution by Large Language Models}, url = {https://arxiv.org/pdf/2305.06311}, year = 2023 }