About me

I’m a 2nd-year PhD student in UCR where I am fortunately advised by Prof. Yue Dong. Prior to that, I received my M.Sc. from Tianjin University (advised by Prof. Deyi Xiong), and my B.S. from Xidian University.

My research interest lies in Natural Language Processing (NLP), specifically conditional text generation, such as summarization and machine translation. Large language models (LLMs) have made significant advancements in conditional text generation, but problems like hallucination and security still exist. Besides these, I also have a strong interest in reinforcement learning (RL). Integrating advanced RL algorithms with NLP is a very interesting and challenging topic, and I look forward to combining them with conditional text generation through more effective and elegant methods.

News

• [10/2024] Completed my internship at Microsoft and really appreciated the guidance from my leaders there. Our paper Not All Heads Matter: A Head-Level KV Cache Compression Method with Integrated Retrieval and Reasoning now available on Arxiv.

• [05/2024] Our paper Safety Alignment in NLP Tasks: Weakly Aligned Summarization as an In-Context Attack was accepted to ACL2024. Demo

• [05/2024] Our paper Cross-Task Defense: Instruction-Tuning LLMs for Content Safety was accepted to NAACL2024 TrustNLP workshop.

• [02/2024] Our paper Safety Alignment in NLP Tasks: Weakly Aligned Summarization as an In-Context Attack now available on Arxiv. Demo

• [12/2023] Our paper Watermarking Conditional Text Generation for AI Detection: Unveiling Challenges and a Semantic-Aware Watermark Remedy was accepted to AAAI2024.

• [10/2023] Our paper Inverse Reinforcement Learning for Text Summarization was accepted to the findings of EMNLP2023. !v!

• [10/2023] Our tutorial on Vulnerabilities of Large Language Models to Adversarial Attacks was accepted to ACL2024.

• [09/2023] I have successfully started my academic journey at UC Riverside. :)

Publications

• Yu Fu, Yufei Li, Wen Xiao, Cong Liu, Yue Dong (2024). Safety Alignment in NLP Tasks: Weakly Aligned Summarization as an In-Context Attack (ACL2024).

• Yu Fu, Wen Xiao, Jia Chen, Jiachen Li, Evangelos Papalexakis, Aichi Chien and Yue Dong (2024).Cross-Task Defense: Instruction-Tuning LLMs for Content Safety (TrustNLP: Fourth Workshop on Trustworthy Natural Language Processing NAACL2024).

• Yu Fu, Deyi Xiong, Yue Dong (2023). Watermarking Conditional Text Generation for AI Detection: Unveiling Challenges and a Semantic-Aware Watermark Remedy (AAAI 2024).
• Yu Fu, Deyi Xiong, Yue Dong (2023). Inverse Reinforcement Learning for Text Summarization (findings of EMNLP2023).
• Jie He, Yu Fu (2023) MetaXCR: Reinforcement-Based Meta-Transfer Learning for Cross-Lingual Commonsense Reasoning (Transfer Learning for Natural Language Processing Workshop).
• Fan Deng, Zhenhua Yu, Wenjing Liu, Xiaoqing Luo, Yu Fu, Ben Qiang, Chaoyang Xu, Zhiwu Li (2021). An efficient policy evaluation engine for XACML policy management (Information Science).

Education

• Ph.D. in Computer Science at University of California, Riverside (2023.9-now)

• M.Sc. in Computer Technology at Tianjin University (2020.9-2023.6)

• B.S. in Software Engineering at Xidian University (2016.9-2020.9)