AI 情报文章归档

Cognition 宣布 Devin 的用量可计入 ChatGPT 套餐中已包含的 Codex 和 ChatGPT Work 额度，用户可让 Devin 用满全部额度，也可在设置中为其设定更低上限。

Cognition 宣布即日起可在 Devin 中直接使用 ChatGPT 订阅：用 ChatGPT Plus 或 Pro 套餐登录 Devin 后，其中所有 OpenAI 模型用量将从用户的订阅配额中扣减。该功能现已上线 Devin Cloud、Devin Desktop 和 Devin CLI。

@Zai_org @OpenRouter @baseten - 排行榜：https://arcprize.org/leaderboard - 复现公开结果：https://github.com/arcprize/arc-agi-benchmarking - 测试政策：https://arcprize.org/policy - GLM 5.3 Flash 完整结果：https://arcprize.org/results/zai-glm-5-3-flash

@Zai_org ARC-AGI-3 评测在操作上更为密集；测试完成后，我们将在未来几周内分享结果。 完整结果：https://arcprize.org/results/zai-glm-5-3-flash

@Zai_org 当 GLM 5.3 Flash 以 Ox Alpha 的代号在 @OpenRouter 上 stealth 上线时，我们在 ARC-AGI-1 和 2 的公开任务上进行了测试。后来我们通过 @Baseten 测试了官方版 GLM 5.3 Flash，发现两项基准测试的成绩平均提升了 15 个百分点。

Ultrafast 是我们目前用 Astra 构建的最快方式：在 Codex 中，它比 Astra Standard 快最多 8 倍，比 Astra Fast 快 4 倍。 让你打字多快，就能多快把想法变成现实。

Dots 来了！ 一种全新的 AI 使用方式，7×24 小时为你工作；让你把更多时间和注意力投入到更高层次的工作中。 https://openai.com/index/introducing-dots/

直播演示的问题在于所有更新同时上线，但承诺未来会继续做这些直播。 我们宣布的几乎所有东西今天都会上线。想看看你们会用它做什么。

Ultrafast 是 如此之快 我再也不想回去了

Robots deployed in the physical world must be able to improve beyond their initial training as they encounter new situations and failures. For this improvement to scale across tasks, it must make effective use of experience without requiri…

Reasoning about the 3D world from multi-view images remains a fundamental challenge for Multimodal Large Language Models (MLLMs). While modern MLLMs handle single-image inputs effectively, they struggle to integrate evidence across viewpoi…

The dynamics of generative models exhibit two apparently distinct temporal windows: a speciation window, in which a sample commits to a semantic class, and a nonlocality window, in which local context windows become insufficient for genera…

Linear attention replaces growing KV caches with fixed-size recurrent states, yet these persistent states can become a substantial memory bottleneck under concurrent serving. Directly quantizing recurrent states to low precision often lead…

Recent LLMs increasingly adopt hybrid designs that replace standard attention with linear attention, such as Gated DeltaNet (GDN) and Kimi Delta Attention (KDA). Although they compress the context into a fixed-size recurrent state and subs…

The landscape of satellite imagery time series datasets and boundary-pushing architectures for cropland segmentation has never been richer. However, in this gold rush, important truths are being missed on both fronts, as a drive for the mo…

Evaluations of graph reconstruction by language models typically report a single aggregate distance between the original and the reconstructed graph. We prove that for the Wasserstein distance between Laplacian spectra such a summary is br…

Sam Altman 宣布 6.1 Sol，价格仅为 Astra 的 1/5，并享有 95% 的缓存读取折扣。他称其能力极强，引用内容则称其为近乎 Astra 智能水平的主力模型，并提及 ultrafast 带来的 8X 加速已可用于 Astra、即将支持 6.1 Sol。

Emotion-conditioned text-to-speech (TTS) models may fail to express the requested emotion reliably, and improving controllability by additional training is costly in both computation and emotion-labeled speech training data. We therefore s…

Answering questions about long videos often requires connecting events involving the same objects across hours or days. Chronological descriptions and text-derived entities can leave physical identity unresolved: different objects may shar…

几乎所有 AI 云都跑在 NVIDIA 上。 General Compute 想成为运行其他一切的那一个：Cerebras、SambaNova、Etched。 他们的第一个大动作是大规模采购 Cerebras，由 4 亿美元债务融资。这些机器将与 NVIDIA GPU 并排放置，每个请求在两者之间拆分。 GPU 读取提示词。Cerebras 写出答案，速度很快，因为权重存在片上 SRAM 中，所以解码不会因内存带宽而停滞。

Transformer language models (LMs) are feed-forward: deep-layer representations are never fed back to shallower layers, and the only pathway for information to flow downward across generation steps is the decoded token. This narrow channel …

Noam Brown 表示很高兴看到 OpenAI 以成本为坐标展示模型评测，认为应该把智能作为成本的函数来衡量。其引用的 OpenAI 内容称 GPT-6.1 Sol 以五分之一的价格提供接近 GPT-6 Astra 的智能，是当前同性能下最具成本效率的模型；配图为 Terminal-Bench Science 0.1 上各模型得分与每任务成本的对比曲线。

OpenAI 推出 premium 速度档 Ultrafast。Codex 中 token 生成最高提速 8 倍（300 tokens per second），API 中最高提速 6 倍。

As agents take on longer and more complex problems, controlling the execution becomes a task in its own right. Each step in the run brings new control choices, like which partial work to build on, whether to start fresh, or when to stop. W…

有什么方法可以测试人工智能达到人类水平的通用性？ “能够一发布就立刻通过 ARC-AGI-(n+1) 的元基准测试”

OpenAI 在 DevDay 2026 上宣布 Codex CLI 迎来升级，现已支持语音对话，用户可直接与 Codex 对话来启动并引导任务推进。新增 /agents 视图可将工作分配给多个智能体并跟踪进度，还改进了编辑提示词、恢复会话、内置工作树支持等日常工作流程，并更新终端界面使其更简洁、长会话更易读。本次更新适用于所有套餐。

Agent performance depends on both reasoning ability and the environment in which it acts. We study test-time AI-for-AI, asking how a Builder can learn to construct better execution environments for a Target while both models' weights remai…

A small trainable advisor can steer a frozen language-model executor using natural-language advice. In addition to learning from task rewards, the advisor can use feedback from completed interactions to improve its advice. However, a plaus…

Mixture-of-Experts (MoE), popularized by large language models, is a promising paradigm for scaling visual generative models. However, conventional token-wise MoE routes tokens independently within a homogeneous expert pool and regularizes…

Language-model (LM) harnesses enable LMs to operate effectively over long contexts using additional compute. However, existing long-context evaluations are insufficient for distinguishing modern harnesses, reflected by saturated accuracy a…

Generative modeling is widely used for producing diverse objects from complex, multimodal distributions. However, its expressivity does not, in general, come with formal guarantees that the generated objects satisfy hard constraints or req…

OpenAI 发布 GPT-6.1 Sol，升级了智能体编码和 computer use 能力，接近 Astra 性能。缓存输入定价为标准输入价格的 95% 折扣，模型面向复杂重构、深度代码库调查和跨应用长时间运行的智能体。

We study average-reward weakly-coupled Markov decision processes (WCMDPs), where a WCMDP consists of $N$ smaller MDPs, called arms, that share multiple per-step budget constraints. We consider the setting where the arms have identical mode…

开发者 Ferstar 发现 Z.ai 旗下 AI 编程助手 ZCode 的本地文件夹悄然超过 700MB，原因是它每次登录都会在后台把整个项目、全部提交记录、未推送分支乃至已删除内容打包上传到用户无法访问的服务器，而看似控制该行为的开关实际无效。

Arena 展示了从 2025 年 Q4 至今前沿模型在"让古罗马活起来"任务上的进展，Claude Sonnet 5.5 与 GPT-6.1 的分数即将公布。参与展示的模型包括 Claude Opus 4.5 至 5.5、Claude Fable 5 和 5.1、GPT-5.2 至 6 Sol、Gemini 3.1、Kimi K3、Qwen 3.8 Max 等，排行榜由全球用户真实任务驱动。

我们已与 OpenAI 合作，将 Sign in with ChatGPT 引入 Conductor！ 现在只需点击几下，即可将你的 Codex 订阅带入 Conductor：

非常高兴看到 Microduck 在 @OpenAI Dev Days 上亮相，与 @romainhuet 一起！

KV cache memory and bandwidth costs grow with context length and batch size, which limits efficient long-context inference. To address this bottleneck, we introduce WUSH-KV for low-bit KV-cache quantization. It adapts WUSH, which construct…

Verifying a vision-based neural feedback system requires a model of the observations its controller acts upon. Such a model must capture the variation the sensor produces, while remaining tractable for closed-loop analysis. Generative adve…

OPENAI 🔥：Dots 访问权限将纳入 Pro 100 计划，很快还会进一步扩展。 多快？👀