智能体生态观察
智能体生态动态速递
追踪 AI 智能体（Agent）生态的模型发布、框架开源、行业落地与安全治理动态。
CERA-MoA: Co-Evolving Routing Mechanisms with Continually Learning LLM Agents
Current Mixture-of-Agents (MoA) paradigms generally treat query routing and agent fine-tuning as separate processes, lim…
ProgramDistill: From Interactive Web Apps to Verifiable Reference-Guided SWE Tasks
Coding agents are typically evaluated with desired behavior specified through issues or instructions. In practical web d…
PANORAMA: Panoptic Grounded Captioning via Mask Proposal Selection
Intelligent systems that act in the world require image understanding that is both comprehensive and spatially grounded.…
ActionPiece: Rethinking Action Tokenization for Autoregressive Vision-Language-Action Models
Action tokenizers play a central role in autoregressive vision-language-action (VLA) models, determining both the target…
PACT: Can Enterprise AI Assistants Be Trusted Under Pressure?
As corporate AI adoption continues to grow, enterprise-grade LLM agents are being deployed into sensitive contexts such …
Evals 实操手册：从 100 条 trace 到一张失败分类表，把误差分析跑成流水线
系列第 ① 篇，对应技能地图第 4 格「评估驱动开发」。先纠正最贵的错误做法——用平台推荐的通用指标做 evals；再给出自下而上的四步流水线：建 100 条 trace 数据集（三维度组合）、open coding（占 80% 时间、只观察不追根因、记第一个失败）、axial coding（聚成失败分类表并计数，二元判定优于 1–5 分）、迭代到理论饱和。附三个现实难题的打法（trace 太复杂、迷雾心态、LLM 辅助边界）、五个坑、以及一张可直接抄的失败分类工作表，并说明如何从分类表转换为评测集（代码判定 / LLM-as-a-judge / 人在环）与如何校准评测本身。
编程智能体 SOP：规划→执行→监控，一张可复制的全链路清单
系列第 ② 篇，对应技能地图第 12–16 格「使用编程智能体」。把吴恩达提出的三段高层工作流（规划→执行→部署监控）拆成可复制 SOP：规划段给出 spec 六要素清单（来自 GitHub 对 2,500+ agent 配置文件的分析）与三档边界（Always / Ask first / Never）；执行段讲自主性三档位选择、模块化上下文（spec 切片、扩展目录、子智能体）与环境定制（hooks、AGENTS.md、裁剪技能）；监控段给出功能/行为/契约三类验证与四个失败模式的对应防护。附 12 步勾选式 SOP 与两条边界提醒（vibe coding ≠ AI 辅助工程；速度/不确定性/成本的致命三角）。
一叶一世界｜什么是 RSI（递归自我改进），什么是 Agent 自进化：一篇读懂
一篇读懂 2026 年最容易被混为一谈的一对概念：RSI（递归自我改进）改进的是自己的"改进能力"，打在权重与 AI 研发流程上、跨用户且不可逆；Agent 自进化不重新训练模型，靠记忆、技能与 harness 让部署后的表现持续变好。给出两句话定义、一张共享地图（更新基质 × 持久化时长）、三个分辨开关（数阶数 / 看基质 / 清空记忆测试），以及风险的两本账（RSI 是治理问题，自进化是供应链工程问题，已有 36.82% 技能含安全缺陷的审计数据）。本文同时为「一叶一世界」栏目开篇。
Anthropic Claude Opus 5.2 灰度测试曝光：响应更快，告别“偷懒”
原标题：《突发！Opus 5.2 深夜上线，RSI 真来了？》 今天清晨，Opus 5.2 悄悄上线了。许多开发者发现，新一代神级模型 Claude Opus 5.2，已经在 Claude Code 中开启灰度测试！ 而且，此前一份内部风险…
豆包工作和飞书，把中国第一个团队 Agent 拉进了工作群
为团队而生的办公 Agent #欢迎关注爱范儿官方微信公众号：爱范儿（微信号：ifanr），更多精彩内容第一时间为您奉上。 ]
荣耀 AgenticOS 亮相，10 月面向 Magic9 系列用户开放尝鲜招募与推送
IT之家 9 月 15 日消息，在今晚的荣耀 HGDC 2026 荣耀开发者大会上，荣耀 AgenticOS 系统也登场亮相。 根据规划， 荣耀年度旗舰 Magic9 系列将首发搭载 MagicOS 11 系统 。今年 10 月份，荣耀还会…
Verifiable Social Reasoning for LLM Assistants
LLM assistants are widely used for daily social advice, yet evaluating their social reasoning in such consultation setti…
Fathom: Per-Query Read Depth for Sparse Decoding over Offloaded KV Caches
When agentic sessions run to a million tokens with many sessions resident at once, the KV cache and the index that ranks…
Emergence World: Adversarial Stress-Testing of Long-Horizon Multi-Agent Systems
As AI agents move from bounded tasks to persistent deployments, failures can propagate through memory, tools, other agen…
Embedding Physics Priors in Robot Learning: A Survey
The rapid progress of artificial intelligence is reshaping robotics and accelerating the adoption of learning-based appr…
Confidence Comes from Experience: Experiential Confidence Estimation from Reasoning to Agents
Reliable confidence estimation is increasingly central to the trustworthy deployment of language models: a calibrated es…
ScienceBuddy: Recursive-in-Recursive Self-Improvement for Interactive Scientific Agents
We introduce and release ScienceBuddy, an interactive scientific research workspace that brings continually improving sc…
Agent-net Open Sources Webagent: A Go Harness That Turns Any Website into a Guarded AI Agent
Agent-net, the team building an agent-to-agent marketplace where AI agents discover, trust, and pay each other, has rele…
EvolveTrade: Experience-Driven Policy Refinement for Self-Evolving LLM Trading Agents
Large language model (LLM) trading agents can combine market data, news, and executable analysis, but their behavior is …
EventEgoHands++: Event-based Egocentric 3D Hand Mesh Reconstruction with Real Dataset
3D hand mesh reconstruction is a challenging yet essential task for downstream applications, including human-robot inter…