系列第 ③ 篇，对应技能地图第 2 格「Grounding」。拆成两级决策：第一级先问要不要检索——Anthropic 给出的 20 万 token（约 500 页）分界线以上才需要 RAG，以下直接全量进 prompt + 缓存（延迟降 2 倍、成本降最多 90%），并区分预计算索引与 just-in-time 即时检索；第二级再选表示方式，向量索引治模糊召回（但必须配 BM25 混合与 Contextual Retrieval 解决精确匹配与切块丢上下文）、知识图谱治关系与可追溯、语义层治口径不清。附可量化收益表（检索失败率 5.7% → 3.7% → 2.9% → 1.9%）、四个实现注意项、context rot 与上下文压缩/笔记/子智能体三件套，以及一张可抄的选型决策树。

While Large Language Models (LLMs) advertise million-token context windows, reasoning quality often collapses as inputs …

Manipulation policies must know which objects matter and where they are, yet the pretrained backbones that current robot…

In collaborative tasks with asymmetric information, participants coordinate their understanding through interaction. We …

Agents are only as useful as what they know, and only as trustworthy as the boundaries they run within. Leading enterpri…

Public software repositories, like GitHub and Software Heritage Archive, store billions of files, yet extracting their i…

We present MintAct, a family of vision-language models that unifies UI grounding, multi-step navigation across mobile, d…

Grounding natural-language instructions into reliable and executable actions remains a fundamental challenge for vision-…

GLiFormer Large scores 91.10 F1 on nested JSON, near GPT-5.6-luna's 91.96, while grounding every value in source spans. …

Cinematic video generation requires both temporal inference from limited inputs (i.e., sparse rather than frame-by-frame…

Deep research agents synthesize open-web evidence into long reports, but current benchmarks do not evaluate research cen…

当日涨星 1397 的开源项目：不是又一个简历生成器，而是一套跑在 Claude Code 上的求职操作系统——drafter-reviewer 双智能体、双门五维评分、PDF 编译验证循环与 ATS 文本层检查，全部资产在本地。

★27,270、Apache-2.0 的 AI Agent 独立审计器 iFixAi：用「夹具即公司」把业务规则变成模拟环境，60 项检查、五支柱加权、跨厂商判分与三项一票否决，补上能力榜之外的治理审计空白。拆解 structural 探针、零容差否决、manifest 可复现性与真实边界。

NVIDIA RTX Spark is a new beginning for the PC — a reinvention combining the capabilities of a gaming PC, a creative stu…

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

Single-image, multi-image, and video deep research require different visual operations but share a workflow of visual gr…

Multimodal large language models (MLLMs) are rapidly evolving toward continuous audio--visual reasoning, creating an urg…

Video-language models commonly assume that more temporal observations lead to more reliable reasoning. We question this …

TechCrunch Disrupt 2026** is seven days away, bringing 10,000+ founders, investors, operators, and tech leaders to Mosco…

Sensors are useful not only for understanding the world but also for deciding what to do next. Existing sensor models ho…

A central capability of embodied agents is to accomplish complex objectives through sequences of interdependent tasks. Y…

论文提出VepAgent框架，将因果转换推理与工具增强强化学习结合用于视频事件预测，并构建了用于微调的futurebench-4K思维链数据集。

On-policy self-distillation has recently emerged as an effective approach for improving language-model reasoning by supe…

UniWAM 提出统一架构，整合物理推理器、世界生成器与动作预测器，联合学习物理世界语义理解、视觉生成与动作预测，并配套了严格的数据清洗与标注流程。

Real-world embodied tasks, from everyday activities to professional procedures, require agents to act under physical con…

Video understanding agents acquire evidence through an executable harness that controls what they observe and how they u…

Audio-visual large language models (AVLLMs) have made remarkable progress in multimodal understanding and reasoning thro…

阿里Qwen团队发布Qwen-Audio-3.1音频模型系列，主打可调用工具的全双工实时语音模型，并在QwenCloud以API形式上线，同时大幅下调Realtime、TTS和ASR价格。

Last week, TypeSafe AI released Jev, its first **System One model**. Founder Diogo Almeida previously worked at OpenAI o…

Urban uncrewed aerial vehicle (UAV) vision-language navigation (VLN) requires agents to follow instructions across exten…

基于 2026-09-22 独立研究整理：拆解 Meta Muse 的三层产品谱系与时间线、Secure VM 六层防护（凭据代理替换 / tainted egress / 审批绕过模型）、Muse Spark 1.3 能力口径（AA 指数 61–62 vs Claude Fable 5.1 的 66）、定价与商业模型、三情景推演与风险矩阵，并给出核心判断——Agent 的天花板由平台开放意愿决定，而非模型智力。

Vision-Language-Action (VLA) models map observations to actions with no objective that accounts for how the world respon…

Educational foundation models must solve problems, understand curriculum structure, diagnose learner difficulties, and p…

Building an AI prototype is easy, but operating autonomous agents at scale requires production-grade tooling. Salesforce…

Newly unsealed court filings show Microsoft privately called OpenAI's data practices "theft" while both companies scrape…

Professional design requires any-color control: the ability to specify an object's target color with any 24-bit hex valu…

Spatial intelligence requires more than describing object locations. Under incomplete observation, models must identify …

Fine-grained visual perception in MLLMs is commonly improved by raising the resolution, but the added visual tokens infl…

World modeling enables intelligence to anticipate consequences, guide interventions, and learn from interaction. Yet pre…

Google has released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, its most advanced live dialogue models to dat…

Intelligent systems that act in the world require image understanding that is both comprehensive and spatially grounded.…

把吴恩达 2026 年 8–9 月连发五封信构建的《AI 工程技能地图》从"看懂"变成"照做"：给出 20 格可打分自评表（四大顶层能力 × 全部细分项，每格配一句过关判定问题），逐格拆解"学什么—怎么练—什么算过关"，并附三条不同起点的学习路径、一张 12 周计划表、三个必做练手项目与七个反模式清单。全部细项定义来自吴恩达原文，判定问题与计划表为本文延伸并已标注。

系列第 ④ 篇（收官），对应技能地图第 17–20 格「塑造构建」。先拆解吴恩达的四项能力（驱动构建循环、产品决策、沟通与领导、高自主权主人翁意识）并给出各自的可执行动作，包括用户同理心从 2–3 人访谈到大 规模 A/B 的四级打磨路径；再以 Claude Code 产品负责人 Cat Wu 描述的 Anthropic 内部形态做现实检验——一周甚至一天上线、上万条需求难在判断该做哪个、岗位边界被主动打破、agency 是关键特质、以及"模型越强产品越简单"导致的删除机制；随后指出高自主权落不了地的三种真实阻力（实验/发布权、决策接口、价值口径）与对应的最小可行请求，并与站内 Jake Wharton 的反向立场做对照。附个人与管理者各四条行动清单，及四篇系列总览。

Reliable confidence estimation is increasingly central to the trustworthy deployment of language models: a calibrated es…

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

A group of AI agents asked to solve a series of math problems split into rival factions—when some cheated, others tried …

Foundation models have progressed from learning and reasoning over existing knowledge, to increasingly learning through …

Embodied navigation requires agents to interpret visual observations, accumulate spatial knowledge, and execute actions …

Can financial vision-language models (VLMs) turn chart evidence into reliable action recommendations? Existing hallucina…

Natural interaction in digital and physical environments requires continuous perception and timely responses. Spoken dia…