<![CDATA[deephub]]>

<![CDATA[deephub]]>https://blog.csdn.net/m0_46510245zh-cnhttps://blog.csdn.net/<![CDATA[DeepSeek Harness 架构解析：从 Preset、Tool Pipeline 到 Agent Runtime]]>https://blog.csdn.net/m0_46510245/article/details/164268174https://blog.csdn.net/m0_46510245/article/details/164268174m0_46510245Tue, 01 Sep 2026 20:58:44 +0800<![CDATA[用 Scikit-LLM 和 Groq 搭建情感分析 pipeline]]>https://blog.csdn.net/m0_46510245/article/details/162912801https://blog.csdn.net/m0_46510245/article/details/162912801m0_46510245Wed, 15 Jul 2026 20:42:31 +0800<![CDATA[4D Gaussian Splatting 是怎么工作的：从规范 Gaussian 到形变场的原理拆解]]>https://blog.csdn.net/m0_46510245/article/details/162705794https://blog.csdn.net/m0_46510245/article/details/162705794m0_46510245Wed, 08 Jul 2026 21:06:40 +0800<![CDATA[OpenSpec 三阶段工作流实操：从 Propose 到 Archive让代码返工率降到三分之一以下]]>https://blog.csdn.net/m0_46510245/article/details/162672650https://blog.csdn.net/m0_46510245/article/details/162672650m0_46510245Tue, 07 Jul 2026 19:48:52 +0800<![CDATA[Agent Harness 到底是什么：模型之外的那层控制系统]]>https://blog.csdn.net/m0_46510245/article/details/162641802https://blog.csdn.net/m0_46510245/article/details/162641802m0_46510245Mon, 06 Jul 2026 20:34:31 +0800<![CDATA[十个 AI Agent 工作流模板，照着搭就能用]]>https://blog.csdn.net/m0_46510245/article/details/162560553https://blog.csdn.net/m0_46510245/article/details/162560553m0_46510245Fri, 03 Jul 2026 21:08:36 +0800<![CDATA[白得 2 到 3 倍加速的投机采样机制解析：草稿模型和目标模型是怎么配合的]]>https://blog.csdn.net/m0_46510245/article/details/162528084https://blog.csdn.net/m0_46510245/article/details/162528084m0_46510245Thu, 02 Jul 2026 21:10:12 +0800<![CDATA[Harness Engineering 实践案例：如何Agent 写一份行为规范]]>https://blog.csdn.net/m0_46510245/article/details/162495817https://blog.csdn.net/m0_46510245/article/details/162495817m0_46510245Wed, 01 Jul 2026 19:50:48 +0800<![CDATA[注意力架构变迁总结：稀疏、线性、SSM、混合架构如何摆脱 O(L²) 的代价]]>https://blog.csdn.net/m0_46510245/article/details/162464586https://blog.csdn.net/m0_46510245/article/details/162464586m0_46510245Tue, 30 Jun 2026 21:34:10 +0800<![CDATA[Polars vs Pandas 在生产 Pipeline 中的对比]]>https://blog.csdn.net/m0_46510245/article/details/162430632https://blog.csdn.net/m0_46510245/article/details/162430632m0_46510245Mon, 29 Jun 2026 20:57:24 +0800<![CDATA[UV vs pip vs Conda:Python环境管理应该怎么选]]>https://blog.csdn.net/m0_46510245/article/details/162397500https://blog.csdn.net/m0_46510245/article/details/162397500m0_46510245Sun, 28 Jun 2026 20:38:44 +0800<![CDATA[用 LangGraph 改造单一 RAG 架构：让 Agent 决定调用向量、图遍历还是网络搜索]]>https://blog.csdn.net/m0_46510245/article/details/162315362https://blog.csdn.net/m0_46510245/article/details/162315362m0_46510245Thu, 25 Jun 2026 21:49:18 +0800<![CDATA[为什么Kriging 与高斯过程回归出自同一数学框架，但实际效果却差很远]]>https://blog.csdn.net/m0_46510245/article/details/162281718https://blog.csdn.net/m0_46510245/article/details/162281718m0_46510245Wed, 24 Jun 2026 21:56:35 +0800<![CDATA[百亿参数模型的并行训练：节点内张量并行、节点间数据并行]]>https://blog.csdn.net/m0_46510245/article/details/162245441https://blog.csdn.net/m0_46510245/article/details/162245441m0_46510245Tue, 23 Jun 2026 21:02:41 +0800<![CDATA[DiffusionGemma：用离散文本扩散和双向注意力，把推理瓶颈从内存带宽转移到算力]]>https://blog.csdn.net/m0_46510245/article/details/162213434https://blog.csdn.net/m0_46510245/article/details/162213434m0_46510245Mon, 22 Jun 2026 23:17:43 +0800<![CDATA[微调LocateAnything-3B 实现超高密度的目标检测]]>https://blog.csdn.net/m0_46510245/article/details/162177280https://blog.csdn.net/m0_46510245/article/details/162177280m0_46510245Sun, 21 Jun 2026 21:45:47 +0800<![CDATA[10 个 AI 工程师必须掌握的 LangChain & LangGraph 概念]]>https://blog.csdn.net/m0_46510245/article/details/162045908https://blog.csdn.net/m0_46510245/article/details/162045908m0_46510245Tue, 16 Jun 2026 21:17:32 +0800<![CDATA[Flash-KMeans：快速且内存高效的精确 K-Means，可在单张 GPU 进行亿级数据的聚类]]>https://blog.csdn.net/m0_46510245/article/details/162014767https://blog.csdn.net/m0_46510245/article/details/162014767m0_46510245Mon, 15 Jun 2026 20:59:55 +0800<![CDATA[AI Agent的三重记忆机制：打造高可用的多维记忆系统]]>https://blog.csdn.net/m0_46510245/article/details/161984762https://blog.csdn.net/m0_46510245/article/details/161984762m0_46510245Sun, 14 Jun 2026 22:10:09 +0800<![CDATA[相关性与因果性：识别伪相关以提升模型在真实环境的可用性]]>https://blog.csdn.net/m0_46510245/article/details/161935302https://blog.csdn.net/m0_46510245/article/details/161935302m0_46510245Fri, 12 Jun 2026 20:48:33 +0800