Agent 工程系统学习第 1 章：从 HTTP 协议层讲透 LLM API——Chat Completions 协议与 role 语义、Function Calling 的"模型选择/代码执行"分工与三大常见错误、SSE 流式手写解析器（含 tool_calls 分块拼接）、token 计量与前缀缓存工程、重试/超时/幂等的错误分类纪律、多模态输入成本。附零框架多轮工具 Agent 实现作业。

LLM 流量与普通 API 流量差异巨大：响应慢且长、按 token 计费、供应商不稳定、还要保留换模型的自由。本文逐项拆解 LLM 网关的核心职责——双层限流、多供应商路由、流式转发、降级链、成本核算与可观测，并给出关键设计要点。

Architect Financial Technologies 发布 Liquid Inference，一个 LLM 推理市场路由器，每次请求通过实时拍卖由各供应商报价竞价，买方支付满足规则中的最低报价，开发者只需替换 base URL 即可让供应商在价格上竞争。

GPT-4 当裁判与人类偏好的一致率 85%，超过了人类彼此之间的 81%——这是 LLM-as-a-Judge 立身的实验，但它交换位置后的一致率只有 65%，一个 token 就能操纵打分，模型版本一漂移榜单就作废。本文拆解机器裁判的奠基实验、三种已知偏差与缓解手段、工程化成本，以及什么时候不该用它。

SSE 是 LLM 应用做打字机效果的主流方案。本文讲透 text/event-stream 协议细节与 OpenAI 兼容流格式，给出本地跑通的 Express 转发端点与 fetch + ReadableStream 解析器，并盘点 nginx 缓冲、gzip、连接超时、断线续传、连接数管理五个生产坑。

LLM assistants are widely used for daily social advice, yet evaluating their social reasoning in such consultation setti…

Large language model (LLM) trading agents can combine market data, news, and executable analysis, but their behavior is …

Skills extend an LLM agent beyond its parametric knowledge, and the gain they promise rests on picking the right one. De…

Large language model (LLM) agents allocate test-time compute adaptively as they revise solutions, use tools, explore alt…

Concurrent local LLM serving on unified-memory desktops must preserve memory headroom and output fidelity, which speed-o…

9 月 11 日科技动态一览：GPT-6 Astra 需求挤爆致 OpenAI 暂停 Pro 20X 新增订阅、Agents API 公测、金融服务版 ChatGPT 上线；Slackbot 升级；加州未成年人社媒法案签署；LG 电视监视争议；观察视角落在"需求侧证实 vs 供给侧反思"的对照上。

Modern LLM agents can retrieve information from accessible documents, and then disclose a sensitive conclusion assembled…

LLM routing dynamically selects and coordinates heterogeneous language models across diverse tasks. However, existing ap…

10 月 18 日，aurora-red/green/blue-alpha 三个无文档、无定价的模型标识短暂出现在 OpenAI API 中随后被移除，测试者推测与图像生成有关。

论文提出AgentSafeEvo，通过攻击者-防御者博弈的协同进化强化学习，提升LLM智能体在多轮工具调用中抵御恶意内容操纵的安全性。

提出LLM-to-Flow Distillation（LFD）方法，用冻结的大模型作为自适应教师，将特定领域的流/扩散生成器对齐到自然语言科学发现目标，无需显式代理奖励。

Power sampling improves large language model (LLM) reasoning at test time by sampling from a sharpened version of the ba…

Developers can build LLM agents by adapting third-party models through benign post-training. We study a supply-chain thr…

Evaluations of LLM systems routinely average over small prompt sets and report models as a ranked table. We ask how much…

Nokia’s applied research team has open-sourced AnyJev, a Python library that turns an open LLM into a decision model. It…

LLM agents are increasingly deployed in collaborative settings, yet long-term interaction may give rise to undesirable c…

We study scheming in LLM agents, in which agents covertly pursue misaligned goals. Our focus is to understand how schemi…

Large language model (LLM) agents increasingly rely on external skills, but routing user requests over large skill regis…

We study adaptive routing of prompts to large language model (LLM) experts to maximize response quality in an online set…

We present a systems-security case study of a two-node split-LLM training system whose privacy evaluation passed while l…

Multi-agent LLM systems commonly use an orchestrator to decompose a task for a team of workers and then improve through …

Quantization has been widely adopted in LLM training and inference to reduce cost and improve efficiency. However, low-b…

AI-powered search products such as ChatGPT search, Google's AI Overviews, and Perplexity provide LLM-synthesized answers…

当日涨星 906 的知识图谱项目：代码用本地 tree-sitter AST 零 LLM 额度建图，文档/图片才走语义趟；52 文件混合语料单次查询 token 降 71.5 倍，LOCOMO recall@10 0.497，每条边带 EXTRACTED/INFERRED/AMBIGUOUS 置信标签。

当日涨星 666 的 Rust 项目：把编码智能体的长期记忆做成 git 版 Markdown wiki（源真相）+ SQLite 派生索引，默认零 LLM 调用，LongMemEval-S 470 题 zero-LLM hit@5 达 0.951，20+ 智能体共享记忆与 claim-once 交接。

★62,973、AGPL-3.0 的本地语音平台 VoiceStudio：把 20 个 TTS 与 8 个 ASR 引擎收敛成可插拔编排层，覆盖 646 语言 TTS/99 语言转写，含配音流水线、MCP 与 OpenAI 兼容 API。拆解引擎子进程隔离、自适应批宽、操作次数性能预算，以及模型许可与 Pro 边界。

!(https://www.gstatic.com/images/branding/googleg/1x/googleg_standard_color_128dp.png)Add as a preferredsource on Google…

传统 APM 盯的是「请求是否 200」，LLM 应用的问题是「200 的请求答得对不对、花了多少钱」。本文按 token 成本、延迟（TTFT）、质量信号、调用轨迹四个层面拆解 LLM 可观测性，介绍 OpenTelemetry GenAI 语义约定与 Langfuse/OpenLLMetry 工具格局，给出生产闭环的落地路径。

VectifyAI 开源的无向量 RAG 引擎 PageIndex（当日涨星 +1,095、★38,082、MIT）：把长文档编译成 JSON 层级树，让 LLM 逐节点推理导航，取代切块+向量相似度。基于它的 Mafin 2.5 在 FinanceBench 全量 10,231 题报告 98.7%。拆解两阶段架构、三模式 TOC 自校验回退、三工具检索循环、成本口径，以及多文档规模化与数据主权的真实边界。

GGUF, GPTQ, AWQ, EXL2, and EXL3 solve the same problem in different ways. This guide separates file containers from quan…

Large language models (LLMs) provide a flexible interface for long-horizon robot planning, but generated plans often fai…

Test-time scaling can improve large language model reasoning by generating and combining multiple candidate responses. I…

Current Mixture-of-Agents (MoA) paradigms generally treat query routing and agent fine-tuning as separate processes, lim…

Safety guardrails in open-weight language models can be readily bypassed using Refusal Feature Ablation (RFA), a techniq…

Backdoor poisoning attacks add poisoned examples to otherwise-clean finetuning data, pairing a trigger with a target beh…

DeepSeek 将 V4.1 Flash 更新至 1017 版，同步向公众开放 Search 与 Fetch 两个接口，定价均为 5 元/千次，按成功调用计费。

Split language models keep input text on a client but send intermediate representations to a server. Attacks can reconst…

针对Rank-GRPO每次训练仅选一条回复导致优势信号主要反映抽样差异的问题，论文提出ReMaRk方法，将生成列表与所有回复标题的并集对比打分并均匀采样请求，在7个骨干模型和3个数据集上以25%-53%的步数达到原方法2000步的验证召回率。

Decoder hidden states from different layers of large language models (LLMs) often encode complementary information. Howe…

2,510 当日涨星的开源 AI 创作 IDE（★7,621、Rust/Tauri）：把 62 个图像/视频/音频/3D 模型收进一个本地桌面端，用 3D 场景、图层合成与人偶摆姿把「提示词抽卡」变成可复现的创作工程。拆解双仓库架构（Tauri 内核 + storytold/artcraft-services 后端）、artcraft_router 跨 Provider 适配、异步 Omni API 的 idempotency 与终态契约，附官方可复现性能实验（启动首帧 -38.1%）与自定义 fair source 许可证风险。

Black-box optimization (BBO) arises in many scientific and engineering problems where objective evaluations are expensiv…

Oct 7, 2026 OpenAI announced two updates to its API platform.** The new Decisions API evaluates text, images, or both, r…

Ollama 负责本地推理，Open WebUI 负责多模型对话、知识库与账号——本文在 Apple M4 上实测 Docker 组合部署全流程：官方命令、RAG 默认参数、OpenAI 兼容 API 的正确开关（v0.11 已改名），以及端口冲突、代理假 IP 导致拉模型失败等真实踩坑清单。

开源权重与闭源 API 是两种供给模式：一个靠生态与自部署取胜，一个靠数据飞轮与前沿能力领跑。本文拆解双方的打法与护城河，解释「落后但够用」的动态平衡，并给出数据敏感性、运维能力、定制需求、成本结构四个维度的选型框架。

选模型 API 的本质是选供给关系，不只是选模型。本文给出官方 API、云托管、聚合网关、自托管四层的对比地图，逐条展开七项评估清单，并给中小团队一条从网关起步的务实路线。