搜索结果：生成式引擎优化

This survey provides a comprehensive framework for optimizing LLM agent workflows through agentic computation graphs (ACGs), distinguishing between static and dynamic methods based on when workflow structure is determined, and proposing structure-aware evaluation metrics beyond traditional task performance. 原文翻译： 本综述通过智能体计算图（ACGs）为优化LLM智能体工作流提供了一个全面的框架，根据工作流结构确定的时间区分静态和动态方法，并提出了超越传统任务性能的结构感知评估指标。

This article explores how large language models (LLMs) function as document completion engines and demonstrates how to build practical applications by mapping between user and document domains, using GitHub Copilot's prompt engineering pipeline as a detailed case study. 原文翻译： 本文探讨了大型语言模型（LLM）如何作为文档完成引擎工作，并通过在用户领域和文档领域之间建立映射来构建实际应用，以GitHub Copilot的提示工程流程作为详细案例研究。

This article explores the most pressing challenges in LLM inference, such as memory bottlenecks and computational inefficiencies, and provides practical solutions including KV caching, batching strategies, and model parallelization techniques using tools like TensorRT-LLM and NVIDIA frameworks. 原文翻译： 本文探讨了LLM推理中最紧迫的挑战，如内存瓶颈和计算效率低下，并提供了实用的解决方案，包括KV缓存、批处理策略和使用TensorRT-LLM和NVIDIA框架的模型并行化技术。

A comprehensive curated collection of learning resources for AI agents, including foundational courses, conceptual guides, framework tutorials, and evaluation benchmarks, regularly updated to help professionals understand, build, and evaluate autonomous AI/LLM agents. 原文翻译： 一份全面精选的AI智能体学习资源集合，包括基础课程、概念指南、框架教程和评估基准，定期更新以帮助专业人士理解、构建和评估自主AI/LLM智能体。

Epsilon is an AI-powered search engine designed for scientific research that scans over 200 million papers to provide answers with citations, search publications and patents, extract information from multiple papers simultaneously, and synthesize research by summarizing and organizing papers into libraries. It's trusted by over 30,000 researchers worldwide but will shut down on April 30, 2026. 原文翻译： Epsilon是一款专为科学研究设计的AI搜索引擎，可扫描超过2亿篇论文，提供带引用的答案、搜索出版物和专利、同时从多篇论文中提取信息，并通过总结和组织论文到库中来综合研究。它被全球超过3万名研究人员信赖，但将于2026年4月30日关闭。

This article analyzes the winning solutions from the Enterprise RAG Challenge, detailing the architectures, models, and experimental approaches used by top performers like Ilya Rice and Emil Shagiev. It provides insights into effective RAG patterns, including dense retrieval, LLM reranking, router patterns, and structured outputs. 原文翻译： 本文分析了企业RAG挑战赛的获胜解决方案，详细介绍了Ilya Rice和Emil Shagiev等顶级选手使用的架构、模型和实验方法。文章深入探讨了有效的RAG模式，包括密集检索、LLM重排序、路由模式和结构化输出。

OCRISp is a one-click RAG implementation tool that combines GUI, CLI, and MCP capabilities into a single executable, allowing users to embed PDFs and connect with MCP clients for data retrieval. 原文翻译： OCRISp 是一个一键式 RAG 实现工具，将 GUI、CLI 和 MCP 功能集成到单个可执行文件中，允许用户嵌入 PDF 文件并通过 MCP 客户端连接进行数据检索。

BlaBlaBlAI is an open-source chat system that enables multiple LLMs and humans to interact in the same room with full awareness of all participants and complete chat history, featuring cost attribution and flexible chat creation. 原文翻译： BlaBlaBlAI是一个开源聊天系统，允许多个LLM和人类在同一个房间内互动，所有参与者都能完全感知彼此并查看完整的聊天历史，具有成本归因和灵活的聊天创建功能。

Librarium is a command-line tool that parallelizes research queries across multiple search and deep-research APIs, collecting and normalizing results into structured output. It features 18 built-in providers across three tiers (deep-research, AI-grounded, raw-search) and supports custom providers. 原文翻译： Librarium 是一个命令行工具，可将研究查询并行分发到多个搜索和深度研究 API，收集并规范化结果，输出结构化数据。它包含 18 个内置提供商，分为三个层级（深度研究、AI 基础搜索、原始搜索），并支持自定义提供商。

This article provides a comprehensive comparison and ranking of over 100 AI models (LLMs) across key metrics including intelligence, price, performance, speed (tokens per second & latency), and context window size. It identifies top performers in each category, such as Gemini 3.1 Pro Preview and GPT-5.4 (xhigh) for intelligence, Mercury 2 and Granite 3.3 8B for speed, and Qwen3.5 0.8B for affordability. 原文翻译： 本文对100多款AI大模型（LLMs）在智能、价格、性能、速度（每秒令牌数及延迟）和上下文窗口大小等关键指标上进行了全面比较和排名。文章识别了各分类中的顶级模型，例如Gemini 3.1 Pro Preview和GPT-5.4 (xhigh)在智能方面表现最佳，Mercury 2和Granite 3.3 8B速度最快，Qwen3.5 0.8B最具性价比。