How Far Can In-Context Alignment Go? Exploring the State of In-Context Alignment
School of Computer Science and Technology, Beijing Institute of Technology(北京理工大学计算机科学与技术学院);Beijing Institute of Technology Southeast Academy of Information Technology(北京理工大学东南信息技术研究院)
Heyan Huang, Yinghao Li, Huashan Sun, Yu Bai, Yang Gao
近期研究表明,In-Context Learning(ICL,上下文内学习)通过使用特定示例,可以使 Large Language Models(LLM,大语言模型)与人类偏好对齐,这被称为 In-Context Alignment(ICA,上下文内对齐),表明模型无需调整参数即可理解人类指令。然而,对 ICA 的机制和适用性的探索仍然有限。本文首先将 ICA 中使用的上下文文本分为三类:格式、系统提示和示例。通过消融实验,我们研究了各部分在使 ICA 有效发挥作用方面的效果。随后,我们考察这些部分的变化如何影响模型的对齐性能。研究发现表明,示例部分对于增强模型的对齐能力至关重要,示例的变化会显著影响对齐性能。我们还全面评估了 ICA 在各类对齐任务中的零样本能力。结果表明,与参数微调方法相比,ICA 在基于知识的任务和工具使用任务中表现出更优性能。但它在多轮对话和指令遵循等方面仍表现出一定局限。
英文摘要:
Recent studies have demonstrated that In-Context Learning (ICL), through the use of specific demonstrations, can align Large Language Models (LLMs) with human preferences known as In-Context Alignment (ICA), indicating that models can comprehend human instructions without requiring parameter adjustments. However, the exploration of the mechanism and applicability of ICA remains limited. In this paper, we begin by dividing the context text used in ICA into three categories: format, system prompt, and example. Through ablation experiments, we investigate the effectiveness of each part in enabling ICA to function effectively. We then examine how variants in these parts impact the model's alignment performance. Our findings indicate that the example part is crucial for enhancing the model's alignment capabilities, with changes in examples significantly affecting alignment performance. We also conduct a comprehensive evaluation of ICA's zero-shot capabilities in various alignment tasks. The results indicate that compared to parameter fine-tuning methods, ICA demonstrates superior performance in knowledge-based tasks and tool-use tasks. However, it still exhibits certain limitations in areas such as multi-turn dialogues and instruction following.
Exploring Multilingual Concepts of Human Value in Large Language Models: Is Value Alignment Consistent, Transferable and Controllable across Languages?
School of New Media and Communication, Tianjin University(天津大学新媒体与传播学院);College of Intelligence and Computing, Tianjin University(天津大学智能与计算学部)
Exploring LLM-based Data Annotation Strategies for Medical Dialogue Preference Alignment
School of Computer Science, Peking University(北京大学计算机学院);Key Laboratory of High Confidence Software Technologies(PKU), MOE, China(教育部高可信软件技术重点实验室(北京大学));Beijing Jiaotong University(北京交通大学)
Exploring Adversarial Robustness and Safety Alignment in Multilingual Multi-Modal Large Language Models
Mohamed Bin Zayed University of AI, UAE(穆罕默德·本·扎耶德人工智能大学,阿联酋);Khalifa University, UAE(卡比拉大学,阿联酋);Australian National University, Australia(澳大利亚国立大学,澳大利亚)
Exploring CLIP's Dense Knowledge for Weakly Supervised Semantic Segmentation
Academy for Engineering and Technology, Fudan University(复旦大学工程技术研究院);Digital Medical Research Center, School of Basic Medical Sciences, Fudan University(复旦大学基础医学院数字医学中心);Shandong Computer Science Center (National Supercomputer Center in Jinan)(山东省计算机科学中心(国家济南超级计算中心))
When Do Model Internals Help? Exploring the Role of Representation Engineering in LLM Safety
State Key Laboratory of Multimedia Information Processing, Peking University(北京大学多媒体信息处理国家重点实验室);School of Computer Science, Peking University(北京大学计算机科学学院)
IBMEA: Exploring Variational Information Bottleneck for Multi-modal Entity Alignment
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空间安全学院)
Through the Lens of Split Vote: Exploring Disagreement, Difficulty and Calibration in Legal Case Outcome Classification
Technical University of Munich(慕尼黑工业大学);Graduate Institute of International and Development Studies(国际与发展研究研究生院);IT University of Copenhagen(哥本哈根IT大学);LMU Munich(慕尼黑大学);Munich Center for Machine Learning (MCML)(慕尼黑机器学习中心);ELTEMATE