AI 情报文章归档

Stop trusting AI edits on faith. Learn how to build post-edit verification hooks that make Claude Code run git diff, tests, and type checks before you ever see the changes, turning agent sessions into self-validating workflows.

Running local LLMs via Ollama, LM Studio, or llama.cpp usually starts with the same guessing game:...

LLM agents are increasingly used for open-ended tasks, very common in research scenarios where they are required to explore, investigate, or act curiously. At inference time, an LLM is a fixed map from context to next-token distribution, t…

Mobile robotic sensing systems must adapt to operator requirements, limited resources, changing connectivity, and observation opportunities shaped by physical motion. Evaluating runtime strategies therefore requires assessing whether their…

Putting AI Agents on Call: How We Built an Autonomous SRE War Room in Slack with Switch and...

AI assistants advise users about diet, exercise, and weight loss---areas in which unsafe guidance can worsen body dissatisfaction, dietary restriction, compulsive exercise, and other behaviors associated with eating disorders (EDs). Existi…

We present a novel GPU-accelerated global solver for constrained nonlinear programming (NLP) inspired by the efficient linear bound propagation framework in neural network (NN) verification. Existing work on NN verification, a problem that…

Diffusion language models (DLMs) generate text through iterative denoising, where decisions about which tokens to commit or remask strongly affect output quality. Existing certainty-aware strategies rely on instantaneous confidence or entr…

Flip the Switch Hackathon 2026 A build log on three agents, one human veto, and what happened when...

The Judge Reads Tokens: An Adversarial Evaluation of an Eval Toolkit's Own...

Background: Well-annotated, diverse medical imaging datasets remain a major constraint for vision-based medical AI. We evaluated Nano Banana, a general-purpose image generation and editing model, as a source of synthetic medical images for…

Building SYNKAI with Switch: a coding agent, an independent verification boundary, and a human...

Reinforcement-learning-trained agents for retrieval-augmented generation retrieve content (passages, knowledge-graph fragments, or hyperedges) and supervise training with outcome rewards on the final answer. Across these paradigms the reas…

I ship as an AI-augmented solo builder, but I never trust what the model produces. Here's the review discipline that caught a missing Android permission in every earlier release build.

We study group-conditional conformal risk control (CRC) for disjoint and overlapping groups under bounded monotone losses. For disjoint groups, we establish exact finite-sample conditional risk control, two-sided bounds on the group-restri…

Local AI promises capable personal assistants that run on users' own devices, keeping data and computation under their control. Open-weight models that fit on local devices can now handle practical tasks such as coding. Yet current inferen…

The carryover problem Most prompt-injection work treats the attack as a one-turn event....

This paper proposes a physics-informed graph neural operator for power system dynamic state estimation. From a sparse, intermittent measurement window at a few generator buses, the operator estimates the rotor-angle and speed trajectories …

The Fast Tree Transform is a locally-aggregated transform defined for arbitrary graph geometries that generalizes both the Hadamard frequency transform and the Haar wavelet transform on the circle. The pure frequency part of the transform …

Machine learning models for omics-based prediction of cancer outcomes are commonly built on aggregated data without distinguishing ancestry groups, or under an ancestry-discrete design, which partitions samples into ancestry groups and fit…

Existing methods enhancing an LLM's long-context reasoning capability often rely on data synthesis and follow-up LLM post-training. However, the data synthesis strategy and the post-training recipes are empirically driven and computational…

Zeroth-order methods provide a way to optimize stochastic nonconvex objectives using only noisy function evaluations. However, the smoothing inherent in their gradient estimators effectively modifies the objective being optimized. In parti…

Large language model (LLM) agents need to choose actions whose consequences unfold over multiple future steps, yet poor decisions do not reveal whether the model failed to represent those consequences or failed to use them. Mechanistic int…

TechCrunch 分析中期选举改变华盛顿权力格局后，硅谷公司试图重建与民主党关系，此前两年它们与特朗普政府建立了更紧密联系。文中引述多位人士观点称 OpenAI 招聘 Jessica Schumer、把数据中心改称 compute campuses 等举动属于公关修复，真正的监管之争可能更多发生在州层面，涉及数据中心选址和就业议题；选举次日还出现万亿美元级市场抛售。

Membership inference attacks (MIAs) aim to determine whether a sample was used during the training of a target model. Strong MIAs rely on reference (or shadow) models to estimate target-conditional score distributions, but training such en…

We study decentralised drone formation, without a central planner, under the conditions that make it hard: nobody knows how many drones are alive, failures are silent, radio is short-ranged and lossy, and actuation is noisy. In our control…

Agentic reinforcement learning (RL) rollouts interleave model reasoning with tool calls in persistent workspaces. Scaling these rollouts incurs substantial memory costs for concurrent environments, while slow tool calls delay entire rollou…

One branch has one step, the other has two, both feed combine. LangGraph.js runs combine twice. Why, and the two fixes: array edges and defer: true.

Model editing provides an efficient way to update factual knowledge in transformer models. Recent approaches reverse edits by modifying, masking, or decomposing model weights, but make recovery dependent on weight-space interventions. We i…

Automatic context compaction and human approval gates are on a collision course. A one-line entry in the OpenAI Agents SDK changelog names the bug, and the fix is an approval record that lives outside the conversation.

Exploratory Data Analysis (EDA) aims to discover informative patterns in large datasets. For transactional data, Association Rule Mining (ARM) is an established approach that facilitates EDA by efficiently generating candidate patterns. Be…

Six AI bot names in 14 days of logs on one site: 105 of 499 requests verified, 176 from outside the operator's IP list. How to check your own.

美国司法部于当地时间 11 月 5 日宣布超级智能工作组（DOJ-SITF）投入运作，统筹涉及 SI 的刑事和民事诉讼，保护创新与防范攻击并行。根据 10 月 13 日备忘录，工作组将协调内部调查、提出执法优先事项并制定起诉及合规评估指引，由斯坦利·伍德沃德和斯蒂芬·韦斯牵头，反垄断局、刑事局、FBI 等单位派代表参与，并对接政府层面的超级智能任务（SIF）。

OpenCode 宣布 OpenCode Browse 正式可用。该产品让智能体控制用户的真实浏览器，可使用现有的标签页、登录状态和扩展，详情见 https://opencode.ai/product/browse。

Prompt injection, hallucinated numbers, runaway costs and leaked card data: practical guardrails for putting AI agents into a security-sensitive platform.

作者反驳 agent-desktop 已解决 AI 操作电脑抢鼠标问题的说法，称在 MacOS 上其后台处理通过率只有 20%，Codex 为 33%，最高的是 Cua Driver 的 58%，即 Cindy 的 Computer Use 采用的方案。被引用内容称 Vercel 开源的 agent-desktop 可通过系统无障碍接口让多个 AI 在后台各用虚拟光标同时操作软件。

A client's internal app (an inventory dashboard connected to 4 live Shopify stores) had been built...

Once you use a coding agent every day, the model itself stops being the main lever. What changes the...

Session files stored fourteen messages while the model saw six. A restart keeps your name, and seven turns later the same conversation forgets it.

Crusoe 确认将怀俄明州东南部一个未完工的数据中心项目出售给 Google，后者计划投资 170 亿美元开发该园区。Google 曾因成本和工期问题放弃成为该项目客户，原项目与 Tallgrass 合作规划最终扩展到 10 吉瓦电力容量。Crusoe 仍在为 Google 开发得州 Amarillo 附近的数据中心，并在 9 月完成 39 亿美元融资。