超越聊天机器人：如何构建智能代理记忆（由 MongoDB 赞助）.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Breakout sessionMikiko ChandrasekharStaff Developer Advocate,MongoDBUnderstanding Agent Memory 202

2、5,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.The Problem“.95%of organizations are getting zero return.The outcomes are so starkly divided across both buyers(enterprises,mid-market,SMBs)and builders(startups,vendors

3、,consultancies)that we call it the GenAI Divide.Just 5%of integrated AI pilots are extracting millions in value”“The core barrier to scaling is not infrastructure,regulation,or talent.It is learning.Most GenAI systems do not retain feedback,adapt to context,or improve over time.”TitleReferenceDescri

4、ptionKey InsightLost in the Middle:How Language Models Use Long ContextsLiu et al.,2024(Stanford/UW)Establishes the fundamental lost in the middle phenomenon in long-context processingModels exhibit U-shaped performance curves with 20%+drops when information is positioned in the middle of long input

5、sContext Rot:How Increasing Input Tokens Impacts LLM PerformanceHong et al.,2025(Chroma)Comprehensive evaluation of 18 LLMs across 194K calls showing performance degradationEven simple tasks models handled perfectly at short lengths fail as input length increasesNoLiMa:Long-Context Evaluation Beyond

6、 Literal MatchingModarressi et al.,2025(ICML)Extends needle-in-haystack tests by removing literal matches between questions and answers72.4%of real-world tasks require external knowledge,revealing benchmark limitationsRULER:Whats the Real Context Size of Your Long-Context Language Models?Hsieh et al