SubmissionNumber#=%=#3 FinalPaperTitle#=%=#UTF: Under-trained Tokens as Fingerprints —— a Novel Approach to LLM Identification ShortPaperTitle#=%=# NumberOfPages#=%=#6 CopyrightSigned#=%=#Jiacheng Cai JobTitle#==# Organization#==#Shanghai Jiao Tong University 800 Dongchuan Road, Shanghai, China 200240 Abstract#==#Fingerprinting large language models (LLMs) is essential for verifying model ownership, ensuring authenticity, and preventing misuse. Traditional fingerprinting methods often require significant computational overhead or white-box verification access. In this paper, we introduce UTF, a novel and efficient approach to fingerprinting LLMs by leveraging under-trained tokens. Under-trained tokens are tokens that the model has not fully learned during its training phase. By utilizing these tokens, we perform supervised fine-tuning to embed specific input-output pairs into the model. This process allows the LLM to produce predetermined outputs when presented with certain inputs, effectively embedding a unique fingerprint. Our method has minimal overhead and impact on model's performance, and does not require white-box access to target model's ownership identification. Compared to existing fingerprinting methods, UTF is also more effective and robust to fine-tuning and random guess. Author{1}{Firstname}#=%=#Jiacheng Author{1}{Lastname}#=%=#Cai Author{1}{Username}#=%=#jccai Author{1}{Email}#=%=#imjccai@gmail.com Author{1}{Affiliation}#=%=#Shanghai Jiao Tong University Author{2}{Firstname}#=%=#Jiahao Author{2}{Lastname}#=%=#Yu Author{2}{Username}#=%=#jiahaoyu Author{2}{Email}#=%=#yujiahao@sjtu.edu.cn Author{2}{Affiliation}#=%=#Shanghai Jiao Tong University Author{3}{Firstname}#=%=#Yangguang Author{3}{Lastname}#=%=#Shao Author{3}{Email}#=%=#shaoyangguang23@mails.ucas.ac.cn Author{3}{Affiliation}#=%=#University of Chinese Academy of Sciences Author{4}{Firstname}#=%=#Yuhang Author{4}{Lastname}#=%=#Wu Author{4}{Email}#=%=#yuhang.wu@northwestern.edu Author{4}{Affiliation}#=%=#Northwestern University Author{5}{Firstname}#=%=#Xinyu Author{5}{Lastname}#=%=#Xing Author{5}{Email}#=%=#xinyu.xing@northwestern.edu Author{5}{Affiliation}#=%=#Northwestern University ========== èéáğö