المؤلفون: Ren, Xuan, Wu, Biao, Liu, Lingqiao

مصطلحات موضوعية: Computer Science - Computation and Language, Computer Science - Artificial Intelligence

الوصف: This paper explores an intriguing observation: fine-tuning a large language model (LLM) with responses generated by a LLM often yields better results than using responses generated by humans. We conduct an in-depth investigation to understand why this occurs. Contrary to the common belief that these instances is simply due to the more detailed nature of LLM-generated content, our study identifies another contributing factor: an LLM is inherently more "familiar" with LLM generated responses. This familiarity is evidenced by lower perplexity before fine-tuning. We design a series of experiments to understand the impact of the "familiarity" and our conclusion reveals that this "familiarity" significantly impacts learning performance. Training with LLM-generated responses not only enhances performance but also helps maintain the model's capabilities in other tasks after fine-tuning on a specific task.

URL الوصول: http://arxiv.org/abs/2402.11192

المؤلفون: REN, Xuan, LIN, Weifang, YI, Jun, LIU, Guojing, TANG, Weijia, YANG, Shuili

المصدر: 2023 IEEE International Conference on Energy Internet (ICEI) ICEI Energy Internet (ICEI), 2023 IEEE International Conference on. :102-106 Oct, 2023

Relation: 2023 IEEE International Conference on Energy Internet (ICEI)

المؤلفون: Han, Qing^{Aff1, Aff2}, Zi, Yun-Jiang^{Aff1, Aff2}, Feng, Tian-Yu^{Aff1, Aff2}, Wu, Nan^{Aff3, Aff4}, Zou, Ren-Xuan^{Aff1, Aff2}, Zhang, Jing-Yu^{Aff1, Aff2}, Zhang, Ru-Lei^{Aff1, Aff2}, Yang, Qing^{Aff3, Aff4, IDs00044024032115_cor8}, Duan, Hong-Xia^{Aff1, Aff2, IDs00044024032115_cor9}

المصدر: Medicinal Chemistry Research. 33(5):740-747

المؤلفون: Ren, Xuan, Liu, Lingqiao

مصطلحات موضوعية: Computer Science - Computation and Language, Computer Science - Artificial Intelligence, Computer Science - Machine Learning

الوصف: Despite significant advancements in existing models, generating text descriptions from structured data input, known as data-to-text generation, remains a challenging task. In this paper, we propose a novel approach that goes beyond traditional one-shot generation methods by introducing a multi-step process consisting of generation, verification, and correction stages. Our approach, VCP(Verification and Correction Prompting), begins with the model generating an initial output. We then proceed to verify the correctness of different aspects of the generated text. The observations from the verification step are converted into a specialized error-indication prompt, which instructs the model to regenerate the output while considering the identified errors. To enhance the model's correction ability, we have developed a carefully designed training procedure. This procedure enables the model to incorporate feedback from the error-indication prompt, resulting in improved output generation. Through experimental results, we demonstrate that our approach effectively reduces slot error rates while maintaining the overall quality of the generated text.

URL الوصول: http://arxiv.org/abs/2306.15933

المؤلفون: Yang, Linyi, Song, Yaoxiao, Ren, Xuan, Lyu, Chenyang, Wang, Yidong, Liu, Lingqiao, Wang, Jindong, Foster, Jennifer, Zhang, Yue

مصطلحات موضوعية: Computer Science - Computation and Language, Computer Science - Artificial Intelligence

الوصف: Machine learning (ML) systems in natural language processing (NLP) face significant challenges in generalizing to out-of-distribution (OOD) data, where the test distribution differs from the training data distribution. This poses important questions about the robustness of NLP models and their high accuracy, which may be artificially inflated due to their underlying sensitivity to systematic biases. Despite these challenges, there is a lack of comprehensive surveys on the generalization challenge from an OOD perspective in text classification. Therefore, this paper aims to fill this gap by presenting the first comprehensive review of recent progress, methods, and evaluations on this topic. We furth discuss the challenges involved and potential future research directions. By providing quick access to existing work, we hope this survey will encourage future research in this area.
Comment: 25 pages, OOD Generalization, Survey

URL الوصول: http://arxiv.org/abs/2305.14104

المؤلفون: Yuan, Yiling, Yao, Shaochen, Ren, Xuan, Xiao, Yao, Zhou, Zhongyu, Lu, Yanfeng

المصدر: 2022 5th International Conference on Data Science and Information Technology (DSIT) Data Science and Information Technology (DSIT), 2022 5th International Conference on. :1-5 Jul, 2022

Relation: 2022 5th International Conference on Data Science and Information Technology (DSIT)

المؤلفون: Ren, Xuan-Qi^Aff11, Nie, Xiao-Hua^Aff11, Chang, Liang^Aff11, Xu, Fei^{Aff11, Aff12}, Guo, Wen-Jie^Aff11

المساهمون: Chinese Society of Aeronautics and Astronautics, Chaari, Fakher, Series Editor^Aff1, Gherardini, Francesco, Series Editor^Aff2, Ivanov, Vitalii, Series Editor^Aff3, Haddar, Mohamed, Series Editor^Aff4, Cavas-Martínez, Francisco, Editorial Board Member^Aff5, di Mare, Francesca, Editorial Board Member^Aff6, Kwon, Young W., Editorial Board Member^Aff7, Trojanowska, Justyna, Editorial Board Member^Aff8, Xu, Jinyang, Editorial Board Member^Aff9

المصدر: Proceedings of the 6th China Aeronautical Science and Technology Conference : Volume III. :277-284

المؤلفون: Da, Tao, Zhang, Dalin, Ren, Xuan, Zhu, Peng, Zhao, Xiaoxu, Wu, Ning, Liu, Xiaohe

المصدر: 2021 IEEE Sustainable Power and Energy Conference (iSPEC) Sustainable Power and Energy Conference (iSPEC), 2021 IEEE. :1916-1921 Dec, 2021

Relation: 2021 IEEE Sustainable Power and Energy Conference (iSPEC)

المؤلفون: Shi, Kun, Ren, Xuan, Da, Tao, Sun, Haixiang, Zou, Dehu, Cheng, Fuming, Liu, Wei

المصدر: 2021 IEEE Sustainable Power and Energy Conference (iSPEC) Sustainable Power and Energy Conference (iSPEC), 2021 IEEE. :4052-4057 Dec, 2021

Relation: 2021 IEEE Sustainable Power and Energy Conference (iSPEC)

المؤلفون: Liu, Kun, Ren, Xuan, Wang, Bin

المصدر: 2021 International Conference on Power System Technology (POWERCON) Power System Technology (POWERCON), 2021 International Conference on. :1811-1815 Dec, 2021

Relation: 2021 International Conference on Power System Technology (POWERCON)