Cross-Refine: Improving Natural Language Explanation Generation by Learning in Tandem. Wang, Q., Anikina, T., Feldhus, N., Ostermann, S., Möller, S., & Schmitt, V. In Rambow, O., Wanner, L., Apidianaki, M., Al-Khalifa, H., Eugenio, B. D., & Schockaert, S., editors, Proceedings of the 31st International Conference on Computational Linguistics, pages 1150–1167, Abu Dhabi, UAE, January, 2025. Association for Computational Linguistics.
Cross-Refine: Improving Natural Language Explanation Generation by Learning in Tandem [link]Paper  abstract   bibtex   
Natural language explanations (NLEs) are vital for elucidating the reasoning behind large language model (LLM) decisions. Many techniques have been developed to generate NLEs using LLMs. However, like humans, LLMs might not always produce optimal NLEs on first attempt. Inspired by human learning processes, we introduce Cross-Refine, which employs role modeling by deploying two LLMs as generator and critic, respectively. The generator outputs a first NLE and then refines this initial explanation using feedback and suggestions provided by the critic. Cross-Refine does not require any supervised training data or additional training. We validate Cross-Refine across three NLP tasks using three state-of-the-art open-source LLMs through automatic and human evaluation. We select Self-Refine (Madaan et al., 2023) as the baseline, which only utilizes self-feedback to refine the explanations. Our findings from automatic evaluation and a user study indicate that Cross-Refine outperforms Self-Refine. Meanwhile, Cross-Refine can perform effectively with less powerful LLMs, whereas Self-Refine only yields strong results with ChatGPT. Additionally, we conduct an ablation study to assess the importance of feedback and suggestions. Both of them play an important role in refining explanations. We further evaluate Cross-Refine on a bilingual dataset in English and German.
@inproceedings{wangCrossRefineImprovingNatural2025,
	address = {Abu Dhabi, UAE},
	title = {Cross-{Refine}: {Improving} {Natural} {Language} {Explanation} {Generation} by {Learning} in {Tandem}},
	shorttitle = {Cross-{Refine}},
	url = {https://aclanthology.org/2025.coling-main.77/},
	abstract = {Natural language explanations (NLEs) are vital for elucidating the reasoning behind large language model (LLM) decisions. Many techniques have been developed to generate NLEs using LLMs. However, like humans, LLMs might not always produce optimal NLEs on first attempt. Inspired by human learning processes, we introduce Cross-Refine, which employs role modeling by deploying two LLMs as generator and critic, respectively. The generator outputs a first NLE and then refines this initial explanation using feedback and suggestions provided by the critic. Cross-Refine does not require any supervised training data or additional training. We validate Cross-Refine across three NLP tasks using three state-of-the-art open-source LLMs through automatic and human evaluation. We select Self-Refine (Madaan et al., 2023) as the baseline, which only utilizes self-feedback to refine the explanations. Our findings from automatic evaluation and a user study indicate that Cross-Refine outperforms Self-Refine. Meanwhile, Cross-Refine can perform effectively with less powerful LLMs, whereas Self-Refine only yields strong results with ChatGPT. Additionally, we conduct an ablation study to assess the importance of feedback and suggestions. Both of them play an important role in refining explanations. We further evaluate Cross-Refine on a bilingual dataset in English and German.},
	urldate = {2025-05-22},
	booktitle = {Proceedings of the 31st {International} {Conference} on {Computational} {Linguistics}},
	publisher = {Association for Computational Linguistics},
	author = {Wang, Qianli and Anikina, Tatiana and Feldhus, Nils and Ostermann, Simon and Möller, Sebastian and Schmitt, Vera},
	editor = {Rambow, Owen and Wanner, Leo and Apidianaki, Marianna and Al-Khalifa, Hend and Eugenio, Barbara Di and Schockaert, Steven},
	month = jan,
	year = {2025},
	pages = {1150--1167},
}
Downloads: 0
About
Documentation
Keywords
Search
Users
Terms and Conditions of Use
Privacy Policy
Sitemap
© BibBase.org 2021