与「AutoML」相关的搜索结果

开源项目 LoopX：超长程 Agent 自主运行 200+ hours，状态不漂移。 我的技术主张是：LLM 上下文有限，长程 Agent 需要外置状态，通过完备的状态管理、监督和规划，让 Agent 无人干预时跑得稳、持续有产出；有人干预时跑得更好，能吸收反馈继续演进。 两条真实 trajectory 分别跨越 220.7 / 272.9 小时，跨多轮执行、等待、人工决策、writeback 与 resume 后，整个 loop 仍能找回目标、证据和下一步。 目前 LoopX 已有 3 个 showcase：auto PR issue fix、AutoML experiment 和 auto coredump fix。 以 OpenViking 开源仓库的 PR issue fix 为例，Agent 不只是循环写代码。它需要持续理解 issue 的不同状态，判断何时开发、何时等待、何时请求 review，处理 CI、冲突和上游变化，并连续交付多个 PR。 这对应 LoopX 的 domain state 管理：领域系统决定真实状态，LoopX 负责把状态投影成下一步可执行的工作。 与此同时，Agent 还可以在干活过程中实现能力自进化。当它发现现有系统缺少某项能力时，可以提出 feature、完成开发与验证、发布新的离线或在线版本，再使用新能力继续原来的任务。 长程 Agent 天然适合自进化，“完成工作”和“升级完成工作的系统”可以在同一条长程轨迹中发生。 LoopX 把这些信息外置成结构化控制面： • Goal / Vision：目标是什么，什么不能被局部优化牺牲 • Todo / Gate：当前执行的 frontier，以及必须留给人的关键判断 • Identity / Authority：谁能 claim、writeback、approve • Evidence / Receipt：每次推进留下什么可回读证据 • Quota / Scheduler：何时继续执行，何时安静等待 • Handoff / Recovery：换模型、换会话、换 host 后如何恢复 你也可以把它理解为一块专门给 Agent 设计的可执行 Kanban。 普通看板只展示“谁在做什么”；LoopX 的状态会直接约束和驱动下一次 bounded turn，让看板本身成为执行系统的一部分。 这套系统最强的地方是通用性。它不只可以修 PR，还可以做 auto research、长期实验、自媒体运营、复杂 feature 开发和办公任务。 Agent 不再只是一次性的回答机器，而可以围绕一个人的 vision，长期工作、等待、吸收反馈、积累证据并持续演进。 LoopX 从一开始 build in public：状态协议、CLI、控制面实现和真实运行轨迹都进入了开源仓库；它也已经和 OpenViking、NoKV 等 agent infra 项目形成了开源合作伙伴关系。 我希望 LoopX 最终能放大每个人的 vision 和想象力。只要你有自己的目标、想法或技术主张，就能拥有一个全天候继续工作和探索的 Agent 系统，帮助你把愿景一点点变成现实。 欢迎试用、提 issue、贡献代码，或者用一个真实的 multi-day task 跑 LoopX。

🧵 Deli AutoResearch SKILL is now officially open source! 🎉 Alongside it, we’re dropping our 4th survey paper — this time on Self-play. Inspired by AlphaZero, we got a powerful insight: prior knowledge doesn’t always lift the ceiling. Models can discover more globally optimal solutions just by playing against themselves. The biggest change in this paper? For the first time, the AutoResearch Agent autonomously planned GPU experiments — and submitted actual RL runs on the DeepSeek 285B model. The entire RL pipeline — experiment design, code writing, running, debugging, and conclusion summarization — was 100% automated, with zero human intervention from me. This was incredibly difficult, but an incredibly important step. GRPO is the tool being called by the AutoResearch Agent here. We see this as the beginning of our Continual Learning research journey. 🚀 As always, this is my personal research project, unaffiliated with any organization. All views are my own. #AI# #ReinforcementLearning# #SelfPlay# #OpenSource# #AutoML# #ContinualLearning# #DeepSeek#

If you're curious about AutoML, check out this book, co-authored by @haifeng_jin from the Keras team. You'll learn about how you can automate your ML pipelines with AutoKeras and KerasTuner!

😳 AUX ÉTATS-UNIS, DES POLICIERS ABUSENT DE LEUR POUVOIR LORS D’UN CONTRÔLE ROUTIER… La situation change complètement lorsqu’ils découvrent que l’automobiliste qu’ils viennent d’interpeller est en réalité un agent du FBI.

This slide from the Anthropic IPO roadshow was a big factor in my decision to leave. It’s an insane plot and I’m surprised they made it public. It shows the compute efficiency of pretraining over time. The y-axis is a **log scale**: they are seeing continued superexponential algorithmic progress. Mythos was not a one-time deviation, but a point on a superexponential. They could already retrain Mythos with a fraction of the compute originally needed. It doesn’t strictly imply a software-only intelligence explosion- they have been spending more on R+D. But algorithmic improvements do not require ridiculous creative genius and are on the verge of being automated. You do the math.

Anthropic lanzó una guía de 37 minutos para construir Agentes de IA que automatizan una empresa entera. Gratis. De los ingenieros que construyeron Claude. Agentes que trabajan, se reparten tareas y ejecutan todo solos. Subtitulado al español.

Great deals are done silently. 🤫 Stealth Bids: make an offer on an NFT you love, and only the owner sees it. No public bidding war to get outbid in. Off-Market Asks: set your asking price privately. Only you see it, and when a bid meets or beats your price, it sells automatically. Explore both on GONDI.

GCSA Insight: Critical Security Vulnerability Discovered in a Leading Robotics Manufacturer’s Product, with System-Level Privilege Compromise Verified on Physical Hardware As artificial intelligence and embodied AI technologies continue to advance, robots are evolving from conventional automated machines into intelligent systems capable of perception, decision-making, and autonomous operation. At the same time, cybersecurity risks are extending beyond software and data into the physical world. The security of robotic systems is no longer solely a matter of protecting data and devices; it can also directly affect operational safety, physical environments, and user safety. Recently, the security research team of the Global Cybersecurity Alliance (GCSA) discovered and verified a critical security vulnerability in a product developed by a leading robotics manufacturer. Through in-depth analysis and multiple rounds of testing on physical hardware, the team successfully breached the device’s system security boundaries and obtained the highest level of operating system privileges in a real-world device environment, demonstrating the vulnerability’s actual impact under specific testing conditions. This research encompassed the entire process, from identifying the security flaw and assessing its risks to validating its impact on physical hardware. The findings also highlight significant challenges facing the underlying security mechanisms of embodied AI systems. The research findings and supporting evidence have been formally submitted to the manufacturer. On November 13, 2026, the manufacturer responded by email, preliminarily confirming the existence of the vulnerability and stating that it had initiated internal remediation efforts and an assessment of the scope of affected products. It is important to emphasize that confirmation of a vulnerability does not mean the associated security risks have been eliminated. GCSA will continue to monitor the manufacturer’s assessment of the affected scope, remediation plans, and subsequent security validation, and will update its research findings as new information becomes available. For intelligent systems capable of physical interaction and movement, timely, effective, and verifiable security remediation is particularly important. In accordance with responsible vulnerability disclosure principles, GCSA is currently withholding sensitive information, including the manufacturer’s identity, specific product details, the vulnerability’s underlying mechanisms, and exploitation methods. GCSA is currently seeking guidance from relevant authorities and regulatory bodies regarding the associated security risks. Taking into account the manufacturer’s remediation progress and a comprehensive assessment of potential security implications, GCSA will carefully determine the appropriate arrangements for further disclosure. Subject to appropriate conditions and upon obtaining the consent of the relevant authorities and regulatory bodies, GCSA will publicly release a detailed security insight report, providing the industry with ongoing updates on the research findings and associated risk assessments. The findings of this research are based on specific testing environments and the evidence obtained to date. They do not imply that other product models or similar robotic systems are affected by the same security issue. This security research and validation on physical hardware represent one of GCSA’s initiatives in the field of embodied AI security. GCSA will continue to advance security research, coordinated vulnerability disclosure, and industry collaboration to promote stronger security response mechanisms across the artificial intelligence sector, enhance product security, and strengthen industry trust. 🔗 Read the full report: · Stay Connected with GCSA Follow GCSA's official social media channels for the latest news, product launches, research developments, and ecosystem updates. GCSA Official Telegram Channel: GCSA Official WeChat Account: #GCSAInsight# #GCSACybersecurity# #GCSASecVibe#

if you want to automerge thousands of pull requests , this is how she does it (via @poteto). 809 in june. 10,926 in october. merged prs a month that touch the grok bot team's code. i went through her posts and put the setup in one place. every pr still gets an agentic risk review plus bugbot. low- and medium-risk ones automerge. only higher-risk ones get a person. day to day, in her words, it's just grok bot and cursor projects. she used to have more cursor automations and replaced them with slack bots that have routines. she works on many streams in parallel with pstack: perf, bug fixes, agent-friendly refactors, her own skills, features, ui polish, the harness, release and observability. not from sitting in a review inbox. she's said the speed gains come from parallelizing. a single chat can take longer because of verification. she handholds it less, so she can run several large projects at once. the real setup is in the codebase and the checks around it. she treats the codebase as memory. every agent mistake needs to turn into a constraint. her team built their own framework for grok bot so some anti-patterns become hard to write at all. agents have to verify their own work. they run the app, and can sometimes model it in tla+. they produce screenshots or video that are easier to check than the whole pr. formal verification skills and ordinary run-and-debug skills work together. she calls this an inverse pyramid. push as much as you can into determinism (fuzzing, property tests, formal and runtime verification, stricter languages) instead of spinning up more agent swarms. in her oct 6 post, she tells sandcastle, a grok team bot, to cut. it runs a buildkite dry run, dms every pr author (they can object), kicks the build, and watches it. then it starts a fuzz swarm: usually 10+ agents on grok 4.7 xhigh, some directed, some chaos monkey, using their verification skills and feature map. when something is high priority, she tells sandcastle to @ mention poteto, her engineer bot, which then spins up a cursor project to triage and fix. her team ships a new stable version of the client at least once a day. that's a lot of machinery under "just grok bot and cursor projects."

Hydrogen brings major leaps in intelligence, speed, and reliability: ⚡ Cutting-edge intelligence: Surpasses Argon on the @ArtificialAnlys Intelligence Index at default high thinking (and climbs even higher at xhigh), while finishing tasks faster than competing models in its class. 💻 Cleaner coding & debugging: More than triples 3.8 Flash on Terminal-Bench 4.1, cuts fabricated claims by more than 5X, and adapts cleanly when a tool call fails or a search returns empty. 💼 Top marks on real-world work: Ranks #1# on Vals Tax Agent Bench, ties for #1# on Vals Legal Research, and leads on AutomationBench-AA, Multimodal Knowledge Agent (MKA), and Agents' Last Exam.