Extremely large pre-trained language models (PTMs) such as GPT-3 are usually released as a service. It allows users to design task-specific prompts to query the PTMs through some black-box APIs. In such a scenario, which we call Language-Model-as-a-Service (LMaaS), the gradients of PTMs are usually unavailable. Can we optimize the task prompts by only accessing the model inference APIs? This paper proposes the black-box tuning framework to optimize the continuous prompt prepended to the input text via derivative-free optimization. Instead of optimizing in the original high-dimensional prompt space, which is intractable for traditional derivative-free optimization, we perform optimization in a randomly generated subspace due to the low intrinsic dimensionality of large PTMs. The experimental results show that the black-box tuning with RoBERTa on a few labeled samples not only significantly outperforms manual prompt and GPT-3's in-context learning, but also surpasses the gradient-based counterparts, i.e., prompt tuning and full model tuning.
Tsinghua University（清华大学）；Zhongguancun Laboratory（中关村实验室）；Institute for AI Industry Research (AIR), Tsinghua University（清华大学人工智能产业研究院（AIR））；Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）
MIST: Jailbreaking Black-box Large Language Models via Iterative Semantic Tuning
School of Computer Science and Information Engineering, Hefei University of Technology（合肥工业大学计算机科学与信息工程学院）；College of Computer Science and Technology, Zhejiang University（浙江大学计算机科学与技术学院）；College of Cyber Security, Jinan University（暨南大学网络安全学院）
CBP-Tuning: Efficient Local Customization for Black-box Large Language Models
Institute of Information Engineering, Chinese Academy of Sciences（中国科学院信息工程研究所）；School of Cyber Security, University of Chinese Academy of Sciences（中国科学院大学网络安全学院）
DISC-LawLLM: Fine-tuning Large Language Models for Intelligent Legal Services
Fudan University（复旦大学）；Huazhong University of Science and Technology（华中科技大学）；New York University Shanghai（上海纽约大学）；Northwest University of Political and Law（西北政法大学）
FedDTPT: Federated Discrete and Transferable Prompt Tuning for Black-Box Large Language Models
China University of Mining and Technology, Beijing（中国矿业大学（北京））；University of British Columbia（不列颠哥伦比亚大学）；University of Texas at Dallas（德克萨斯大学达拉斯分校）；The Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳））
Discrete Prompt Tuning via Recursive Utilization of Black-box Multimodal Large Language Model for Personalized Visual Emotion Recognition
Graduate School of Information Science（信息科学研究生学校）；Technology, Hokkaido University, Sapporo 060-0814, Japan（技术，北海道大学，札幌060-0814，日本）；Office of Institutional Research, Hokkaido University, Sapporo 060-0808, Japan（机构研究办公室，北海道大学，札幌060-0808，日本）；Faculty of Information Science（信息科学学院）
On Active Privacy Auditing in Supervised Fine-tuning for White-Box Language Models
University of Chinese Academy of Sciences（中国科学院大学）；Nanjing Artificial Intelligence Research of AI（南京人工智能研究院）；Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）
Privacy-Preserving Parameter-Efficient Fine-Tuning for Large Language Model Services
University of Ottawa（渥太华大学）；Zhongguancun Laboratory（中关村实验室）；School of Electrical Engineering and Computer Science（电气工程与计算机科学学院）；Department of Computer Science and Technology（计算机科学与技术系）；Institute for AI Industry Research (AIR)（人工智能产业研究院）
Black-box Model Ensembling for Textual and Visual Question Answering via Information Fusion
Faculty of Computer Science, University of Vienna（维也纳大学计算机科学学院）；UniVie Doctoral School Computer Science, University of Vienna（维也纳大学UniVie计算机科学博士学院）；Department of Computer Science, Aarhus University（奥胡斯大学计算机科学系）；Faculty of Philological and Cultural Studies, University of Vienna（维也纳大学语文与文化研究学院）
Mitigating Fine-tuning based Jailbreak Attack with Backdoor Enhanced Safety Alignment
University of Wisconsin-Madison（威斯康星大学麦迪逊分校）；University of Michigan-Ann Arbor（密歇根大学安娜堡分校）；Princeton University（普林斯顿大学）；University of California, Davis（加州大学戴维斯分校）；University of Chicago（芝加哥大学）
NEFMind: Parameter-Efficient Fine-Tuning of Open-Source LLMs for Telecom APIs Automation
Alto University, School of Science, Espoo, Finland（阿尔托大学科学学院，芬兰）；Networked Systems Security (NSS) Group -- KTH Royal Institute of Technology, Stockholm, Sweden（网络系统安全（NSS）小组——皇家理工学院，瑞典）；Ericsson, Finland（爱立信，芬兰）
To See is Not to Learn: Protecting Multimodal Data from Unauthorized Fine-Tuning of Large Vision-Language Model
School of Computing；Augmented Intelligence, Arizona State University, Tempe, AZ, USA；Department of Computer Science；Engineering, Texas A\&M University, College Station, TX, USA
Gnothi Seauton: Empowering Faithful Self-Interpretability in Black-Box Transformers
School of Artificial Intelligence, Shanghai Jiao Tong University（上海交通大学人工智能学院）；Shanghai Jiao Tong University（上海交通大学）；Sichuan University（四川大学）；Big Data and AI Lab, ICBC（中国工商银行大数据与人工智能实验室）；Hong Kong University of Science and Technology, Guangzhou（香港科技大学广州校区）
Mix and Match: Learning-free Controllable Text Generation using Energy Language Models
University of California San Diego（加州大学圣迭戈分校）；Toyota Technological Institute at Chicago (TTIC)（芝加哥丰田技术研究院（TTIC））
控制文本生成的最新工作要么需要基于属性的微调的基础语言模型（LM），或限制了参数化的属性映射，以兼容的基础自回归LM。在这项工作中，我们提出了Mix and Match LM，这是一种基于全局分数的可控文本生成方法，它结合了任意预训练的黑盒模型，用于在生成的文本中实现所需的属性，而不涉及任何关于黑盒模型的微调或结构假设。我们将可控生成的任务解释为从基于能量的模型中抽取样本，该模型的能量值是来自黑盒模型的分数的线性组合，黑盒模型分别负责流畅性、控制属性和对任何条件上下文的忠诚度。我们使用Metropolis-Hastings抽样方案从这个基于能量的模型中使用双向上下文和全局属性特征进行抽样。我们验证了我们的方法在各种控制生成和基于风格的文本修订任务的有效性，优于最近提出的方法，涉及额外的训练，微调，或限制性的假设模型的形式。
RTD-Guard: A Black-Box Textual Adversarial Detection Framework via Replacement Token Detection
Institute of Information Engineering, Chinese Academy of Sciences（中国科学院信息工程研究所）；School of Cyber Security, University of Chinese Academy of Sciences（中国科学院大学网络安全学院）
