AI 情报文章归档

推出 Realtime Director。 这款新工具可让你实时导演高质量视频生成。 由 @fal 的 H3 Max 提供支持。

Image tokenizers define the ``visual language'' of unified multimodal models, yet are commonly studied through isolated metrics or generation-/understanding-only evaluations. These evaluations do not fully capture how visual tokens behave …

The digitization of healthcare has generated vast, longitudinal, and multimodal patient records over a lifetime, yet fully exploiting these data to represent and predict patient state trajectories remains a critical challenge. Current AI m…

Robotic Process Automation (RPA) is widely used to reduce administrative burden in United States hospitals, yet an estimated 30-50% of RPA initiatives underperform because processes are selected informally, without a repeatable method to c…

Existing methods for test-time reinforcement learning (TTRL) derive rewards from answer-level self-voting on unlabeled test-time tasks with canonical answers, but this breaks down for code generation because programs cannot be compared by …

GitHub 上的 i-have-adhd 项目提供一个 skill，用于防止编码智能体把最终答案深埋在输出里。该项目经由 Hacker News 热门讨论传播，HN 获得 101 分。

Agent harnesses (the system prompt, tool set, execution hooks, and context-management scaffolding around a model) are a critical determinant of agentic task success. Automated harness evolution can enable smaller models to perform well on …

Execution feedback can guide coding agents toward correct repository repairs, but only when the tests capture the behavior requested by the issue. Agent-generated tests can encode incomplete or incorrect behavioral targets; when the same t…

An input may activate few hidden units even when different inputs collectively use an entire network. We study the statistical complexity of this input-dependent sparsity in the one-hidden-layer ReLU model of Awasthi et al. (COLT 2024). Fo…

Amari's contributions to information geometry and machine learning are well known. Here, we revisit Amari's work on Bayesian duality which has not received as much attention. We connect Amari's Bayesian duality to a convex duality of Bayes…

Nvidia 和 AMD 越来越多地借助自身资产负债表竞争，通过提供信用担保帮助客户获得融资、建成新的数据中心。The Information 援引一位数据中心高管称，两家有时会为给特定数据中心客户提供信用担保而展开竞价战，Nvidia 因资产负债表更强而更为激进。数据中心即使有土地、电力、GPU 和租户，若出租方或贷款方不相信租户能可靠支付 15 年长约，仍难以获得融资。

Visual encoders construct a representation of the image input for Vision-Language models. How much conceptual, as opposed to immediately visible, information does this representation contain? We use canonical color as a controlled test cas…

Cohere CEO Aidan Gomez 称从多个大实验室员工处听说，消费级 AI 工具的生产用户数据被转化为合成数据用于训练，处理复杂数学、商业、软件、生物问题等模型最需要学习的用例更易被上采样。

Dexterous manipulation involves contact-rich and fine-grained interactions with the physical world, posing significant challenges for existing vision-language-action (VLA) models due to severe visual occlusions and complex contact dynamics…

Noam Brown 转发 Sebastien Bubeck 的澄清并回应称，Navier-Stokes 结果不依赖 Levent 或 Tristan 的提示词，所用模型远超今天的 LLM，并附上 GPT-6 Astra 与内部模型在开放数学题上 pass rate 的对比图。

作者转发 OpenAI 官方声明：OpenAI 称其 AI 求解了悬置约 90 年的 Navier-Stokes 千禧年大奖问题。一个训练中的新内部模型（据称显著强于 GPT-6 Astra）驱动约 10，000 个并发智能体，88 小时得出解，GPT-6 Astra 再用 17 小时完成 Lean 形式化与验证。

Every developer today relies on LLMs. Whether it's drafting boilerplate, refactoring complex logic,...

Normalization renders large parts of neural networks effectively scale invariant, inducing a hidden feedback loop in which learning-rate schedules and weight decay interact through the parameter norm to control the effective step taken by …

Long horizon Large Language Model (LLM) agents rely on external memory systems to preserve user preferences and task knowledge across extended interactions. Conventional retrieval mechanisms optimize semantic compatibility rather than down…

While research on recursive self-improvement (RSI) has predominantly automated model training pipelines, reliable autonomous development demands a missing pillar: post-hoc monitoring and auditing to understand what models learn and ensure …

pstack 0.15.0 发布，主要改进为 token 使用效率提升，根据会话中加载的 skills 数量可节省 3% 至 11%，并清理了多余分号、破折号及冗长表述。新增"攻击前提"和"测试行为而非实现"两条原则，以减少无用测试。使用 /poteto-mode 时这些原则会自动加载，也可通过 /principle-… 按需调用。

过去七天内出现了九份令人不安的新报告，指向 OpenAI 存在严重不当行为模式。这些报告涉及的具体事件与细节尚未在原文中展开，但整体趋势表明该公司的行为问题正在持续累积。文章以此为依据，对 OpenAI 的运营方式提出质疑。

Emad Mostaque 评论 OpenAI 公布的 Navier-Stokes 求解进展，指出报告显示该模型仅从 8 月 28 日开始训练，现已能在几分钟内攻克难题。他表示数学（并希望很快是物理）正被 bitter lesson 改写，并引用他人观点称此前未解决 Millennium Problems 可能是因为没投入大量推理算力。

Nathan Lambert 转发 OpenAI 的演示并评论称，希望看到关于智能体数量的 scaling laws。OpenAI 称其内部模型组用约 10，000 个协作 AI Agent 在 88 小时内得出 Navier-Stokes 解，并全程保留了监控与隔离等前沿评估安全防护。他认为智能体集群数量的扩展是否与单模型 CoT 推理时计算的扩展同等关键，关系到许多关于 AI 未来的讨论。

Curriculum learning is governed by several coupled design choices---how difficulty is defined, how examples are ordered, how much exposure each level receives, and how quickly training moves across levels---making it hard to isolate what a…

Uno 提出一种不改变输出分布的 LLM 加速方案：保留原自回归模型负责质量，仅用扩散模型并行草拟多个 token 供其验证，无需独立草稿模型。在 Qwen3-8B 上，最大测试批次下请求吞吐量提升 2.5 倍，系统吞吐量提升 1.6 倍，端到端 RL 训练提速最高 40%。

Combining search with function approximation has driven major advances in game-playing programs, making self-play algorithms more competitive than ever. Still, the computational overhead of the most popular methods, based on Monte Carlo Tr…

Jason Liu 转发并调侃 OpenAI 的公告，称该模型玩 RuneScape 也很强。引用内容称 OpenAI 分享了对 Navier-Stokes 千禧年大奖难题的一个解法，证明由一组智能体使用比 GPT-6 Astra 能力强得多的 OpenAI 下一代模型产出；该问题关乎 Navier-Stokes 方程建模的光滑三维流体运动描述是否会崩溃，约 90 年未获解决。

Emad Mostaque 转发 OpenAI 的公告并评论称这是迈向 ASI 的里程碑事件。OpenAI 宣布由一组智能体使用显著强于 GPT-6 Astra 的下一代模型，给出了 Navier-Stokes 千禧年大奖难题的一个 blow up 解，该问题涉及三维光滑流体运动的描述是否会崩溃，约 90 年未解。

Large language models (LLMs) may abandon correct positions when users push back, exhibiting a failure mode known as sycophancy. Existing evaluations typically use short, pre-specified conversations and may therefore miss failures that emer…

Large Language Models (LLMs) often exhibit "Attention Sink" (AS) and the accompanying "Massive Activations" (MAs) at the initial position of a sequence. These phenomena frequently co-occur, and MAs can pose challenges for low-bit quantizat…

NYU 数学教授 Tristan Buckmaster 与 Anthropic 数学家 Levent Alpöge 公布针对 Navier-Stokes 存在与光滑性这一千禧年难题的三项证明初步结果，并称其研究进展信息被泄露给 OpenAI，对方随后用大量算力沿其独特路线追赶证明。

Open vocabulary 3D semantic segmentation methods typically lift CLIP features into 3D. This embeds points in a joint vision-language space known to behave like a bag-of-words on compositional tasks. Furthermore, even annotation free varian…

Mid-training, the stage between pre-training and alignment, is where a model's per-domain data composition is typically set by data availability rather than principled design. We ask what that decision buys, and whether a later alignment p…

As LLMs are increasingly used for pre-submission self-review, there is growing demand for feedback that not only identifies weaknesses but also guides authors toward concrete revisions. We study this as Actionable Peer-review Generation an…

In reinforcement learning with verifiable rewards (RLVR) trained with group relative policy optimization (GRPO), the KL-free reward-advantage term studied here depends on within-group reward variation. If all rollouts in a group are correc…

GPT-6 Astra 能快速构建整个场景，但细节不够。 Meshy 7 专为丰富的几何细节而打造。🤩 我们通过 Meshy API 重新生成了薄弱部分，并在相同位置替换。现在它在任何镜头距离下都能撑住。

OpenAI 宣布给出 Navier-Stokes 千禧年大奖难题的一个解，证明由一组智能体使用比 GPT-6 Astra 能力显著更强的 OpenAI 下一代模型完成，该问题关于 Navier-Stokes 方程描述的三维光滑流体运动是否会失去正则性，约 90 年未获解决。Ethan Mollick 转述称这是大事，但学术署名归属与证明竞赛中的时间先后争议仍需解决。

Anthropic 的 Claude Max 计划（每月 100-200 美元）订阅者提起扩大集体诉讼，指控其宣传的 5x、20x 用量上限实为受周上限约束的 5 小时会话额度，误导用户。诉讼由两位曾任职 FTC 的律师代理，7 月首次提交后撤回并重新提交；Anthropic 在撤案动议中称相关说明可通过点击链接获取，公司未回应置评请求。

High-quality tool-use data is critical for training language models to interact effectively with external tools. However, existing synthetic approaches typically follow a generate-then-filter paradigm with static post-hoc verification, oft…