Yafu Li

Yafu Li (李雅夫)

Hello! My name is Yafu Li, and I am a third-year PhD student under joint training of Zhejiang University and Westlake University, under the supervision of Prof. Yue Zhang. I am currently conducting my internship at Tencent AI Lab, where I am mentored by Dr. Leyang Cui and Dr. Wei Bi.

I earned my Bachelor’s degree from Wuhan University, followed by a Master’s degree from the University of Edinburgh, supervised by Prof. Alex Lascarides. Prior to my PhD, I worked as an NLP researcher in Noah Ark's lab at Huawei, under the mentorship of Dr. Liangyou Li and Prof. Qun Liu.

Email / CV(En/中文) / Google Scholar / Semantic Scholar / Twitter / Github

Research

My research focuses on machine translation and natural language generation, with a recent focus on LLM-related topics.. You can find our recent work on detecting AI-generated texts at DeepfakeTextDetect.

	Deepfake Text Detection in the Wild Yafu Li, Qintong Li, Leyang Cui, Wei Bi, Longyue Wang, Linyi Yang, Shuming Shi, Yue Zhang preprint project page / paper link We present a comprehensive benchmark dataset designed to assess the proficiency of deepfake detectors amidst real-world scenarios.
	Explicit Syntactic Guidance for Neural Text Generation Yafu Li, Leyang Cui, Jianhao Yan, Yongjing Yin, Wei Bi, Shuming Shi, Yue Zhang ACL, 2023, Best Paper Nomination (1.6%) project page / paper link We propose a syntax-guided generation schema, which generates the sequence guided by a constituency parse tree in a top-down direction.
	GLUE-X: Evaluating Natural Language Understanding Models from an Out-of-distribution Generalization Perspective Linyi Yang, Shuibai Zhang, Libo Qin, Yafu Li, Yidong Wang, Hanmeng Liu, Jindong Wang, Xing Xie, Yue Zhang ACL Findings, 2023 project page / paper link We present the first attempt at creating a unified benchmark named GLUE-X for evaluating OOD robustness in NLP models
	Consistency Regularization Training for Compositional Generalization Yongjing Yin, Jiali Zeng, Yafu Li, Fandong Meng, Jie Zhou, Yue Zhang ACL, 2023 project page / paper link We propose to boost compositional generalization of neural models through consistency regularization training.
	Revisiting Cross-Lingual Summarization: A Corpus-based Study and A New Benchmark with Improved Annotation Yuleng Chen, Huajian Zhang, Yijie Zhou, Xuefeng Bai, Yueguan Wang, Ming Zhong, Jianhao Yan, Yafu Li, Judy Li, Michael Zhu, Yue Zhang ACL, 2023 project page / paper link We propose ConvSumX, a cross-lingual conversation summarization benchmark, through a new annotation schema that explicitly considers source input context.
	Multi-Granularity Optimization for Non-Autoregressive Translation Yafu Li, Leyang Cui, Yongjing Yin, Yue Zhang EMNLP, 2022 project page / paper link We propose multi-granularity optimization for non-autoregressive translation, which collects model behaviors on translation segments of various granularities and integrates feedback for backpropagation.
	Categorizing Semantic Representations for Neural Machine Translation Yongjing Yin, Yafu Li, Fandong Meng, Jie Zhou, Yue Zhang COLING, 2022 paper link Learning of semantics of atoms and compositions can be improved by introducing categorization to the source contextualized representations.
	Prompt-Driven Neural Machine Translation Yafu Li, Yongjing Yin, Jing Li, Yue Zhang ACL Fidings, 2022 project page / paper link Versatile prompts can be effectively integrated into one single translation model.
	Label Attention Network for Structured Prediction Leyang Cui^, Yafu Li^, Yue Zhang TASLP, * equal contribution paper link We extend Label attention network (LAN) to general sequence labeling tasks including non-autoregressive translation.
	On Compositional Generalization of Neural Machine Translation Yafu Li, Yongjing Yin, Yulong Chen, Yue Zhang ACL, 2021 project page / paper link Neural machine translation suffers poor compositionality.
	Sentence-State LSTMs For Sequence-to-Sequence Learning Xuefeng Bai, Yafu Li, Zhirui Zhang, Mingzhou Xu, Boxing Chen, Weihua Luo, Derek Wong, Yue Zhang NLPCC, 2021 paper link An alternative sequence-to-sequence model architecture obtains comparative performance as the Transformer.