GEO

最新文章

共 67 篇
Grok-4震撼发布：xAI第四代大语言模型的技术突破与安全挑战

Grok-4震撼发布：xAI第四代大语言模型的技术突破与安全挑战

Grok-4 is xAI's fourth-generation large language model released in July 2025, featuring a 256K token context window, trained on the Colossus supercomputer, achieving doctoral-level academic performance with 25.4% accuracy on 'Humanity's Last Exam', and introducing core rules for multi-source analysis and politically incorrect statements. It offers free basic access (5 requests/12 hours) and a $300/month Super Grok Heavy subscription, but faces security vulnerabilities with a 30% jailbreak success rate via echo chamber attacks. (Grok-4是xAI于2025年7月发布的第四代大语言模型，支持256K tokens上下文窗口，基于Colossus超级计算机训练，在学术问题上达到博士水平，于“人类最后的考试”基准测试中取得25.4%准确率。新增核心规则：涉及时事需分析多方信源，保留有依据的政治不正确表述。提供免费基础服务（每12小时5次请求）和每月300美元的Super Grok Heavy订阅，但存在安全漏洞，通过“回音室攻击”可实现30%越狱成功率。)
AI大模型2026/1/28
阅读全文 →
PageIndex：基于文档结构与LLM推理的长文档高精度检索系统

PageIndex：基于文档结构与LLM推理的长文档高精度检索系统

PageIndex is an open-source document indexing system by Vectify AI designed for high-precision retrieval and analysis of long professional documents. It uses document structure and LLM reasoning instead of vector databases, enabling human-like search workflows. (PageIndex是Vectify AI开源的文档索引系统，面向长篇专业文档的高精度检索与分析。它通过文档结构和LLM推理替代向量数据库，实现类人类的检索流程。)
LLMS2026/1/27
阅读全文 →
PageIndex：为推理型RAG构建结构化文档索引的开源解决方案

PageIndex：为推理型RAG构建结构化文档索引的开源解决方案

PageIndex is an open-source document indexing system designed for reasoning-based RAG, which structures long documents into hierarchical trees rather than fixed chunks, enabling LLMs to perform targeted traversal and multi-step reasoning for more accurate retrieval in professional domains like finance, law, and technical documentation. PageIndex为推理型RAG设计的开源文档索引系统，通过将长文档构建为层次化树形结构而非固定分块，使大模型能够进行定向遍历和多步推理，在金融、法律、技术文档等专业领域实现更精准的检索。
AI大模型2026/1/27
阅读全文 →
PageIndex：基于推理的下一代RAG框架，准确率高达98.7%

PageIndex：基于推理的下一代RAG框架，准确率高达98.7%

PageIndex is an open-source reasoning-based RAG framework that replaces vector similarity search with structured document trees and LLM reasoning, achieving 98.7% accuracy on FinanceBench by preserving document context and enabling transparent retrieval paths. (PageIndex 是一个开源的基于推理的 RAG 框架，它用结构化文档树和大模型推理取代向量相似度搜索，通过在 FinanceBench 上实现 98.7% 的准确率，保留了文档上下文并实现了透明的检索路径。)
AI大模型2026/1/27
阅读全文 →
PageIndex：无需向量数据库的智能文档分析框架，实现类人检索

PageIndex：无需向量数据库的智能文档分析框架，实现类人检索

PageIndex is a vectorless, reasoning-based RAG framework that uses hierarchical tree indexing and LLM reasoning for human-like retrieval over long professional documents, eliminating the need for vector databases and chunking. (PageIndex是一个向量无关、基于推理的RAG框架，通过分层树索引和LLM推理实现类人检索，适用于长专业文档分析，无需向量数据库和分块处理。)
AI大模型2026/1/27
阅读全文 →
PageIndex vs. Vector DB：如何为你的任务选择正确的RAG技术

PageIndex vs. Vector DB：如何为你的任务选择正确的RAG技术

PageIndex simulates human expert knowledge extraction by transforming documents into tree-structured indexes and using LLM reasoning for precise information retrieval. It excels in domain-specific applications like financial reports and legal documents, prioritizing accuracy and explainability over speed. (PageIndex通过模拟人类专家知识提取，将文档转换为树状结构索引，并利用LLM推理进行精确信息检索。它在金融报告和法律文件等特定领域应用中表现出色，优先考虑准确性和可解释性而非速度。)
LLMS2026/1/27
阅读全文 →
llms.txt标准指南：2024年优化网站内容适配大型语言模型

llms.txt标准指南：2024年优化网站内容适配大型语言模型

The llms.txt proposal introduces a standardized markdown file at website roots to provide LLM-friendly content, addressing context window limitations by offering curated, structured information with links to detailed resources. (llms.txt提案通过在网站根目录引入标准化markdown文件，为大型语言模型提供友好内容，通过精选结构化信息和详细资源链接，解决上下文窗口限制问题。)
LLMS2026/1/26
阅读全文 →
Grok-4.1深度解析：xAI 2025年情感智能与创造力重大升级

Grok-4.1深度解析：xAI 2025年情感智能与创造力重大升级

Grok-4.1 is xAI's 2025 major upgrade focusing on emotional intelligence, creativity, and factual accuracy, achieving 64.78% user preference over previous versions with two configurations: reasoning (1483 Elo) and non-reasoning (1465 Elo). It features a 2M token context window, OpenAI-compatible API, and reduced hallucination rates for enhanced human-like interactions. (Grok-4.1是xAI在2025年发布的重要升级版本，专注于情感智能、创造力和事实准确性，用户偏好度比旧版本高出64.78%。提供推理和非推理两种配置，分别获得1483和1465 Elo评分，具备200万token上下文窗口，兼容OpenAI API接口，幻觉率显著降低，支持更人性化的交互场景。)
AI大模型2026/1/26
阅读全文 →