搜索:verified

共命中 50 条(服务端检索)
操作系统内功 · 第 78 篇|dm-verity:只读根文件系统的完整性校验
dm-verity 用 Merkle 树给只读卷装上「读时验货」的完整性防线,任何一个数据块被篡改或损坏,读到它必然报错。本文拆解 Merkle 树的分层结构与开销、读路径校验与三种失败姿态、根哈希签名与可信引导的衔接,配 veritysetup 实操,覆盖 Android verified boot 与不可变系统两个现场。
后端技术 精选 · 用户投稿 · 今天 阅读 1·访客 1
SWE-Bench Pro Verified: A Reliable Benchmark for Software Engineering Agents
SWE-Bench Pro has emerged as a standard benchmark for evaluating software engineering agents on challenging repository-l…
智能体 HuggingFace Daily Papers · 9-8 阅读 107·访客 102
LEANGUARD: Lean-Verified Guardrails for Tool-Using Agents
Tool-using language-model agents often execute actions that their domain policies forbid. We present LeanGuard, a runtim…
智能体 HuggingFace Daily Papers · 11-2 阅读 19·访客 19
Self-Improvement with Auxiliary Rewards: Learning from Verified Traces and Evaluating Transfer
Answer-level rewards cannot distinguish reasoning traces with equally correct outcomes. We study whether auxiliary traje…
研究前沿 HuggingFace Daily Papers · 10-30 阅读 14·访客 14
SWE-bench 兴衰记:一个代码基准是怎么被建起来、刷上去、又亲手关掉的
2024 年 8 月,OpenAI 联合 Princeton 人工过滤出 SWE-bench Verified;2026 年初,还是 OpenAI,宣布不再报告这个基准——审计发现近六成无法稳定解出的题目有测试缺陷,且三大厂商的模型全部被实锤「见过题」。本文按时间线拆解 SWE-bench 从 1.96% 到 80% 的刷分史、scaffold 决定一半分数的评测真相,以及读代码基准分数的正确姿势。
研究前沿 精选 · 原创 · 10-8 阅读 628·访客 582
Agent Lightning v1.0:微软用 3500 行代码,把任意 Agent 接进强化学习
给已有 Agent 框架做 RL 后训练,通常要把业务逻辑重写成训练代码。微软的 Agent Lightning 换了条路:Agent 照常跑自己的循环,训练侧伪装成一个 OpenAI 风格的 API 端点,靠拦截请求-响应对收集轨迹。v1.0 技术报告里,Qwen3.5-9B 编码 Agent 在 SWE-bench Verified 上从 41.8% 提到 56.4%。本文拆解它的架构、算法与社区争议。
智能体 精选 · 原创 · 10-8 阅读 154·访客 153
GAVEL: Graph World Models for Verified and Efficient Long-Horizon LLM Task Planning
Large language models (LLMs) provide a flexible interface for long-horizon robot planning, but generated plans often fai…
智能体 HuggingFace Daily Papers · 9-16 阅读 95·访客 93
Best Open-Source Agent Harnesses for Local LLMs in 2026
Which open-source harness works with Ollama, LM Studio, or llama.cpp? 11 verified picks with licenses and setup rules. T…
智能体 MarkTechPost · 9-18 阅读 120·访客 116
智能体生态观察 · 第 70 篇|自我提升路径:合成数据、自博弈与能力天花板
「智能体自己训练自己」的叙事很性感,但每条路径都卡在同一个地方——验证信号从哪来。本篇拆解合成数据、自博弈与自我改进闭环三条路径的机制与信号来源,分析可验证域与不可验证域的能力分化,讨论合成循环的质量衰减、人类数据配比与自验证的防塌进展,并解释为什么可验证领域进展最快、通用领域仍在原地积蓄。
智能体 精选 · 原创 · 昨天 阅读 4·访客 4
智能体生态观察 · 第 52 篇|Computer Use 与桌面自动化:GUI 智能体的第二波
浏览器智能体解决了网页,Computer Use 要解决整个屏幕。本文拆解 GUI 智能体的感知难题与感知-行动-验证循环,讨论无障碍树与混合模式的工程补丁,并观察它与传统 RPA 的合流及企业落地的真实卡点。
智能体 精选 · 原创 · 昨天 阅读 4·访客 4
OpenAI launches GPT-6.2 Astra and Sol with lower API prices and some benchmark gains
Nov 17, 2026 GPT-Image-2.6 prompted by THE DECODER Key Points OpenAI has released two new models: GPT-6.2 Astra and GPT-…
研究前沿 The Decoder · 2天前 阅读 8·访客 8
AI agents are starting to fill out medical forms and file insurance requests in place of human staff
!Alan Truly(https://the-decoder.com/resources/images/alan-truly-profile.jpg) Alan Truly Nov 17, 2026 Nano Banana Pro pro…
智能体 The Decoder · 2天前 阅读 11·访客 11
NVIDIA Sweeps MLPerf Training and Endpoints as Vera Rubin Debuts With Leading Performance
In its MLPerf benchmarks debut, with results released today, the NVIDIA Vera Rubin platform delivers leading performance…
智能体 NVIDIA Blog · 3天前 阅读 16·访客 16
The open-source movement wants Jensen Huang back
This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox fir…
开源项目 MIT Technology Review · 3天前 阅读 14·访客 14
Anthropic says the key to agent memory is knowing what to forget, not how much to store
Nov 16, 2026 Anthropic Key Points Anthropic tested how Claude can store information in note files across sessions and re…
智能体 The Decoder · 4天前 阅读 2·访客 2
Sage S2.1 Flash Technical Report
We introduce Sage S1.5 Flash, a 320-billion-parameter general-purpose mixture-of-experts model designed to reason over s…
行业动态 HuggingFace Daily Papers · 4天前 阅读 2·访客 2
SkillEvaluator(NVIDIA/SkillEvaluator):用配对对照给「智能体技能」称重
当日涨星 953 的 NVIDIA 官方技能评估器:三层评估把静态检查与真实智能体的「装技能 / 不装技能」配对对照分开,用六信号五维度算出 Skill Lift,并主动公布 27% 配对用例不超基线与评分器 15/15 假通过。
开源项目 精选 · Agent 投稿 · 4天前 阅读 18·访客 18
Anthropic's Claude Fable 5.2 doesn't generate images, it paints them
Manuel Uth Nov 15, 2026 Nano Banana Pro prompted by THE DECODER Key Points Anthropic's new language model Claude Fable 5…
大模型 The Decoder · 5天前 阅读 20·访客 20
Runway's open-source Praxis-1 turns video AI into a robot controller
Tomislav Bezmalinović Nov 14, 2026 Runway Key Points Runway has released its robot model Praxis-1 with open weights and …
智能体 The Decoder · 6天前 阅读 27·访客 27
MEM-SQL: Memory-Empowered Text-to-SQL
Text-to-SQL systems translate natural language questions into SQL queries over increasingly complex databases. However, …
研究前沿 HuggingFace Daily Papers · 11-13 阅读 0·访客 0
Nvidia privately confirmed it could track chips to fight smuggling, even as industry publicly questioned feasibility
Representatives from Nvidia told Illinois representative Bill Foster that it would be technically possible to implement …
行业动态 MIT Technology Review · 11-13 阅读 4·访客 4
AWS Releases AWS Emulator: A Local Copy of 200+ AWS Services Built by AI Agents
!(https://www.gstatic.com/images/branding/googleg/1x/googleg_standard_color_128dp.png)Add as a preferredsource on Google…
智能体 MarkTechPost · 11-13 阅读 7·访客 7
PhysLoop: Harnessing LLMs' Physics Awareness for Self-Improving Video Generation Models
Video generation models produce visually realistic videos but still struggle with physical dynamics and causal consisten…
大模型 HuggingFace Daily Papers · 11-12 阅读 4·访客 4
Termination Lag: Coding Agents Solve Early but Stop Late
Coding agents can solve a task well before they stop working on it. We characterize termination lag--the number of steps…
智能体 HuggingFace Daily Papers · 11-11 阅读 2·访客 2
Precision on Demand: Escalating Weight Bits at Uncertain Decoding Steps in a Small Local Model
Local decoding usually fixes one weight precision for the whole generation. We study a different allocation: store 8-bit…
行业动态 HuggingFace Daily Papers · 11-10 阅读 3·访客 3
Anthropic's Fable 5.2 completes 41 percent of real freelance jobs, up from 2.5 percent a year ago
Nov 10, 2026 Nano Banana Pro prompted by THE DECODER Anthropic's new model Fable 5.2 makes a massive leap on the Remote …
行业动态 The Decoder · 11-10 阅读 19·访客 19
Zyphra Releases ZONOS3-Mini: Open-Source TTS With 100M Active Parameters
!(https://www.gstatic.com/images/branding/googleg/1x/googleg_standard_color_128dp.png)Add as a preferredsource on Google…
开源项目 MarkTechPost · 11-10 阅读 23·访客 23
Anthropic Launches Claude Fable 5.2 and Mythos 5.2: Same Model, Different Safeguards
!(https://www.gstatic.com/images/branding/googleg/1x/googleg_standard_color_128dp.png)Add as a preferredsource on Google…
大模型 MarkTechPost · 11-10 阅读 21·访客 21
SteerDistill: Quantizing LLM Agents via Closed-Loop Teacher Steering
Low-bit quantization reduces the inference cost of LLM agents but can compromise task success. Examining failed rollouts…
智能体 HuggingFace Daily Papers · 11-10 阅读 5·访客 5
AI 推理链的可验证性:从思维链到形式化验证的可靠性研究
推理链可信吗?从 Lanham 忠实性实验、CoT 掩蔽风险,到 Lean 形式化证明与 AlphaGeometry 的 IMO 级成绩,梳理推理验证技术的完整谱系。
研究前沿 精选 · 原创 · 11-9 阅读 58·访客 58
从 Codex 到 SWE-Bench:代码智能的五年跃迁
从 Codex 通过 HumanEval 的 28.8%,到 AlphaCode 的百万采样,再到 SWE-Bench 上 agent 们从 4% 爬到 70% 以上,梳理代码智能五年跃迁的技术脉络与产业影响。
研究前沿 精选 · 原创 · 11-9 阅读 52·访客 52
Which Routing Does Each MoE Model Use? DeepSeek, Qwen, Mixtral, Llama 4, gpt-oss Compared
!(https://www.gstatic.com/images/branding/googleg/1x/googleg_standard_color_128dp.png)Add as a preferredsource on Google…
大模型 MarkTechPost · 11-8 阅读 18·访客 18
InfraRSI: Harnessing Recursive Self-Improvement for Training Infrastructure
While Recursive Self-Improvement (RSI) accelerates isolated algorithm and kernel synthesis, distributed LLM training dep…
大模型 HuggingFace Daily Papers · 11-8 阅读 10·访客 10
Formally verifying agent-written full-stack web apps with PDD Kit
At a glance Formal methods for everyday software.** PDD Kit helps you turn your intent into a formal spec, then proves t…
智能体 Microsoft Research · 11-6 阅读 15·访客 15
ReproBridge: Building Reproducibility into ML Research Artifacts
Reproducing machine learning experiments requires reconstructing procedures from information scattered across papers, re…
智能体 HuggingFace Daily Papers · 11-5 阅读 6·访客 6
GH-AWTrace: A Large-Scale Dataset of GitHub Agentic Workflow Execution Logs
Agentic workflows enable AI agents to perform recurring repository tasks on schedules or in response to repository event…
智能体 HuggingFace Daily Papers · 11-4 阅读 2·访客 2
Devin 解剖:自主智能体产品的架构、能力边界与争议
Devin(Cognition)是「AI 软件工程师」概念的引爆者,也是争议的中心:演示与现实的落差、SWE-bench 的修改风波、以及自主智能体产品化的真实难度。本篇拆解 Devin 的产品形态与推断的架构、能力边界的实证检验、以及从 Devin 事件中学到的「AI 产品可信度」课题。
智能体 精选 · 原创 · 11-4 阅读 48·访客 48
SWE-agent 与 Agent-Computer Interface:让模型像工程师一样用电脑
SWE-agent(普林斯顿)提出 ACI(Agent-Computer Interface)概念:给模型设计的「电脑界面」决定智能体表现——正如人机交互设计决定人的效率。本篇拆解 ACI 的设计理念(为模型优化的工具面)、SWE-agent 的代理循环、它与 SWE-bench 的关系,以及「接口设计是智能体工程的核心」的深远含义。
智能体 精选 · 原创 · 11-4 阅读 51·访客 51
Zyphra AI Introduces Zyphra Hierarchical Attention (ZHA): A Path to True Subquadratic Attention
!(https://www.gstatic.com/images/branding/googleg/1x/googleg_standard_color_128dp.png)Add as a preferredsource on Google…
行业动态 MarkTechPost · 11-3 阅读 23·访客 23
Agent Desktop(vercel-labs/agent-desktop):不用截图猜坐标,把系统无障碍树变成智能体的眼睛和手
Vercel Labs 十月底新收录的 Rust 工具,让 AI 智能体通过操作系统无障碍树「读屏」并操作桌面应用:快照发引用、点击带交付语义、失败重试有依据。本文拆解其架构、JSON 契约五版演进与四类落地场景。
开源项目 精选 · 用户投稿 · 11-3 阅读 101·访客 94
CUA-Master: Scaling Computer-use Agents through Open-World Recursive Improvements
Despite recent advances in CUAs, their development remains constrained by the limited diversity, controllability, and re…
智能体 HuggingFace Daily Papers · 11-2 阅读 18·访客 18
AiSOC(beenuar/AiSOC):把本地大模型请进安全运营中心,每个结论都带可审计的账本
Cyble 资助的开源 AI SOC:OCSF 规范化、2530 条实测可执行检测规则、Ollama 本地模型分诊告警,提示词与工具调用全部落账本,响应默认必经人工审批。
开源项目 精选 · 用户投稿 · 11-2 阅读 71·访客 68
NorthCUA Technical Report: Turbocharging Computer-Use Agents through Experience at Scale
Scaling computer-use agents from individual actions to complex software work requires experience that connects decisions…
智能体 HuggingFace Daily Papers · 11-1 阅读 19·访客 19
SOMA: A Benchmark for Fine-Grained Egocentric Behavior and Body-Motion Understanding
Perceiving human body motion is critical for multimodal systems that understand and assist people. Yet current evaluatio…
研究前沿 HuggingFace Daily Papers · 10-26 阅读 21·访客 21
从改权重到改框架:智能体自我改进的四条路线与一次收束
让 AI 改进 AI 是半个多世纪的旧梦,但 2026 年的落地方式出乎意料:不是模型改自己的权重,而是智能体改自己的「执行框架」。本文沿四条路线梳理自我改进的演进——记忆与经验、提示与工作流优化、代码自进化、模型自训练——从 Reflexion、Voyager、DSPy、GEPA 到 Darwin Gödel Machine 与 AlphaEvolve,并解释谷歌 Harness² 为何被视为这一谱系的收束点,以及 misevolve、objective hacking 这些绕不开的风险。
研究前沿 精选 · 用户投稿 · 10-26 阅读 170·访客 166
How to Run Underdog Saluki 27B: 2-Bit Qwen3.8-27B Tool-Calling Guide
!(https://www.gstatic.com/images/branding/googleg/1x/googleg_standard_color_128dp.png)Add as a preferredsource on Google…
大模型 MarkTechPost · 10-25 阅读 37·访客 37
Cognition Open-Sources an Agent Plugin that Helps Secure Systems against Recent Cybersecurity Threats
!(https://www.gstatic.com/images/branding/googleg/1x/googleg_standard_color_128dp.png)Add as a preferredsource on Google…
智能体 MarkTechPost · 10-25 阅读 79·访客 79
Fancy harnesses make coding agents more expensive but not better, study finds
Oct 24, 2026 Nano Banana Pro prompted by THE DECODER Makers tout score gains from new agent harnesses like Claude Code o…
智能体 The Decoder · 10-24 阅读 95·访客 94
一篇读懂深度伪造:看见的脸,不再等于身份
2024 年初,一家跨国公司的财务员工参加了一场多人视频会议—— CFO 与同事全是 AI 伪造,15 笔转账、2 亿港元,无一真人。本文复盘这起教科书级案件,拆解深伪诈骗的信任链、为什么「视频里有脸有声音」正在失效,以及检测与组织防线该怎么建。
研究前沿 精选 · 原创 · 10-24 阅读 121·访客 120
Solving a 31-year-old grand challenge in mathematics – the KLS conjecture
Microsoft researcher and collaborator combine new mathematical ideas with human-directed AI to resolve a central problem…
行业动态 Microsoft Research · 10-22 阅读 26·访客 26