Papers

Preprints

2026

1. Preprint

   Generative AI floods and dilutes the market for books

   Tuhin Chakrabarty, Xinyue Liu, Jane C. Ginsburg, and Paramveer Dhillon

   2026

   arXiv SSRN

   In the news: The Wall Street Journal · Inc. · Business Insider · The New York Times · +1 more

   @article{chakrabarty2026generative,
     title = {Generative AI floods and dilutes the market for books},
     author = {Chakrabarty, Tuhin and Liu, Xinyue and Ginsburg, Jane C. and Dhillon, Paramveer},
     year = {2026},
   }

Publications

2026

1. COLM

   

   Alignment Whack-a-Mole: Finetuning Activates Verbatim Recall of Copyrighted Books in Large Language Models

   Xinyue Liu, Niloofar Mireshghallah, Jane C. Ginsburg, and Tuhin Chakrabarty

   In Third Conference on Language Modeling, 2026

   Oral at the ICML 2026 Workshop on the Impact of Memorization on Trustworthy Foundation Models

   arXiv SSRN Code Website

   In the news: SSRN · The Telegraph · Hacker News

   @inproceedings{liu2025alignment,
     title = {Alignment Whack-a-Mole: Finetuning Activates Verbatim Recall of Copyrighted Books in Large Language Models},
     author = {Liu, Xinyue and Mireshghallah, Niloofar and Ginsburg, Jane C. and Chakrabarty, Tuhin},
     booktitle = {Third Conference on Language Modeling},
     year = {2026},
   }

2025

1. COLM

   

   NoveltyBench: Evaluating Creativity and Diversity in Language Models

   Yiming Zhang, Harshita Diddee, Susan Holm, Hanchen Liu, Xinyue Liu, and 3 more authors

   In Second Conference on Language Modeling, 2025

   HTML Code Website

   @inproceedings{zhang2025noveltybench,
     title = {NoveltyBench: Evaluating Creativity and Diversity in Language Models},
     author = {Zhang, Yiming and Diddee, Harshita and Holm, Susan and Liu, Hanchen and Liu, Xinyue and Samuel, Vinay and Wang, Barry and Ippolito, Daphne},
     booktitle = {Second Conference on Language Modeling},
     url = {https://openreview.net/forum?id=XZm1ekzERf},
     year = {2025},
   }

2024

1. INLG

   

   Customizing Large Language Model Generation Style using Parameter-Efficient Finetuning

   Xinyue Liu, Harshita Diddee, and Daphne Ippolito

   In Proceedings of the 17th International Natural Language Generation Conference, 2024

   arXiv Code Website

   One-size-fits-all large language models (LLMs) are increasingly being used to help people with their writing. However, the style these models are trained to write in may not suit all users or use cases. LLMs would be more useful as writing assistants if their idiolect could be customized to match each user. In this paper, we explore whether parameter-efficient finetuning (PEFT) with Low-Rank Adaptation can effectively guide the style of LLM generations. We use this method to customize LLaMA-2 to ten different authors and show that the generated text has lexical, syntactic, and surface alignment with the target author but struggles with content memorization. Our findings highlight the potential of PEFT to support efficient, user-level customization of LLMs.

   @inproceedings{liu-etal-2024-customizing-large,
     title = {Customizing Large Language Model Generation Style using Parameter-Efficient Finetuning},
     author = {Liu, Xinyue and Diddee, Harshita and Ippolito, Daphne},
     editor = {Mahamood, Saad and Minh, Nguyen Le and Ippolito, Daphne},
     booktitle = {Proceedings of the 17th International Natural Language Generation Conference},
     year = {2024},
     address = {Tokyo, Japan},
     publisher = {Association for Computational Linguistics},
     url = {https://aclanthology.org/2024.inlg-main.34/},
     pages = {412--426},
   }