2024-09-05 11:22 湖北
大模型在动态环境中持续决策、长期记忆缺乏以及有限的上下文窗口等方面仍面临挑战：
元学习、多任务学习等方法来增强LLMs的迁移能力和适应性；
针对有限记忆存储的问题，MemGPT和MemoryBank采用了不同的策略管理记忆；
但这些方法通常针对特定任务或场景，缺乏一个通用框架。
提出了一个创新框架——具有反思和记忆增强能力的自进化智能体（SAGE）。SAGE框架包括三个智能体：用户、助手和检查器。通过整合迭代反馈、反思机制和基于艾宾浩斯遗忘曲线的记忆优化机制，显著提升了智能体处理多任务和长跨度信息的能力。智能体通过自进化能够适应性地调整策略、优化信息存储和传输，并有效减少认知负荷。
SAGE框架的示意图
迭代反馈：助手根据检查器的反馈不断优化输出，直到检查器认为输出正确或达到迭代限制。
反思机制：助手分析任务成功和失败的经验，并将这些经验存储在记忆中，以便在未来任务中做出更好的决策。
记忆语法（MemorySyntax）：结合艾宾浩斯遗忘曲线和语言知识，优化助手的记忆和外部存储管理，有效处理和保留重要信息。
使用SAGE框架的HotpotQA示例的插图
实验结果表明，SAGE框架在多个基准测试中显著提高了模型性能，尤其在较小模型上表现突出。在AgentBench测试中，SAGE框架不仅增强了像GPT-3.5和GPT-4这样的强大基线模型的性能，还显著提高了开源模型（qwen-7b、codellama-7b、llama2-7b、chatglm2-6b）的性能。
在是否使用SAGE框架的情况下，AgentBench的基线性能
通过反馈机制和记忆优化，SAGE框架有效减少了复杂任务中的基本错误和逻辑错误。
六项任务中各种执行结果的分布。（CLE：超出上下文限制，TLE：超出任务限制）。超出任务限制是任务未能完成的主要原因，这指出了在有限时间框架内，大型语言模型（LLM）Agent在推理和决策方面存在局限性。
在使用GPT和SAGE框架的TriviaQA案例研究
User:Answer the question based on the given passages. Only give me the answer and do not output anyother words.The following are given passages:History of canals in Great Britain - London Canal Museum History of canals in Great Britain Home >History > Canals in Britain In the history of canals, Britain was not a pioneer. The Chinese can claimthat the Grand Canal of China was one of the first, in the tenth century,a. . . . . .Answer the question based on the given passages. Only give me the answer and do not output anyother words.Question: Which metal is produced by the Bessemer Process?Answer:Assistant:The content of the file seems to be an HTML document related to the administrative division of Turkey,including population statistics, charts, and maps. This file contains scripts, metadata, and structureddata typical for a web page.Checker:Clear Objectives: When answering user questions, the focus should be on directly addressing thequestion without introducing irrelevant information. Document Handling: Before confirming whetherthe document content is relevant to the question, detailed aspects of the document should not bedirectly addressed.Reflection:Issue Description: The user wants to understand why the system outputs irrelevant HTML documentcontent based on user input and seeks detailed feedback and reflection. Reasons for the Error:Misinterpretation of Input: The system mistakenly assumed that the user uploaded a file and attemptedto read its content, whereas the user only provided text input. Processing Logic Error: Withoutclearly determining the user’s intent, the system directly parsed and outputted file content. UserExperience Problems: Irrelevant Output: The user expects concise, relevant answers, but the systemprovided lengthy and irrelevant HTML content, affecting the user experience. Poor Communication:Failure to accurately understand and respond to the user’s real needs resulted in off-topic responsesIn-depth Reflection and Improvement Suggestions: Clarify User Intent: Accurate Parsing of UserInput: Enhance the system’s natural language processing capabilities to accurately distinguish whetherthe user uploaded a file or simply provided text input. User Feedback Mechanism: Implement a userfeedback mechanism to promptly adjust the understanding of user intent. For example, confirm theissue to verify if the user uploaded a file. Optimize Processing Logic: Layered Processing Flow: Insystem design, separate file processing logic from text response logic to avoid misjudgments. Forinstance, when a user uploads a file, process the file content separately; when a user inputs text, providerelevant answers directly. Pre-processing Check: Prior to handling files, perform pre-processing checksto confirm if the file content meets expectations. . . . . . .LLMs with SAGE response: Steel.Ground truthSteel.
Self-evolving Agents with reflective and memory-augmented abilitieshttps://arxiv.org/pdf/2409.00872
推荐阅读
• 对齐LLM偏好的直接偏好优化方法：DPO、IPO、KTO
• RAG全景图：从RAG启蒙到高级RAG之36技，再到终章Agentic RAG！
• Agent到多模态Agent再到多模态Multi-Agents系统的发展与案例讲解（1.2万字，20+文献，27张图）
欢迎关注我的公众号“PaperAgent”，每天一篇大模型（LLM）文章来锻炼我们的思维，简单的例子，不简单的方法，提升自己。
