2026-07-28

Kimi K3: Open Frontier Intelligence

👍 231

07/27 08:00

We introduce Kimi K3, a 2.8T parameter Mixture-of-Experts model with 104 billion activated parameters, native vision capabilities, and a 1-million-token context window. Kimi K3 is built on Kimi Delta Attention and Attention Residuals, which improve information flow across sequence length and model d

中文介绍 提出Kimi K3，2.8T参数MoE模型（104B激活参数），原生视觉能力与百万token上下文窗口。基于Kimi Delta Attention和Attention Residuals改进序列与模型维度信息流，实现前沿开放智能。

JarvisHub: An Open Harness for Canvas-Native Multimodal Creative Agents

👍 99

07/26 08:00

Creative AI is moving from single-step asset generation toward long-horizon multimodal production. Although recent generative models can synthesize high-quality images, videos, audio clips, UI elements, storyboards, slides, and other creative assets, real-world creative work requires more than isola

中文介绍 提出JarvisHub，面向多模态长周期创意生成的开放式画布原生代理框架。整合图像、视频、音频、UI等资产生成，支持多步协同创作，推动创意AI从单步生成迈向复杂生产。

From Proprietary to Open-Source: Bridging the Distribution Gap via Multi-Agent Protocol Distillation in Agentic Search

👍 63

07/27 08:00

Agentic search enables large language models to solve knowledge-intensive tasks by interleaving multi-step reasoning with retrieval, yet optimizing this with outcome-based reinforcement learning (RL) provides only sparse supervision. Knowledge distillation can supply denser guidance, and advanced pr

中文介绍 针对智能体搜索中闭源到开源模型的分布差异，提出多智能体协议蒸馏（MAPD）。利用高级私有模型的行为协议指导开源模型，弥合分布差距，提升检索-推理联合性能，无需依赖稀疏的强化学习奖励。

StateAct: Program State, before Pixels, for Long-Horizon Computer-Use Agents

👍 50

07/24 08:00

Computer-use agents are usually improved by strengthening perception: better models for reading a screenshot and choosing where to click. Yet a screenshot is only a lossy rendering of the underlying program state, e.g., the files, application backends, and DOM that hold the task data. Different stat

中文介绍 提出StateAct，强调在长时计算机操作代理中利用程序状态（文件、DOM等）而非像素截图。通过访问底层状态信息，比纯视觉感知更高效、可靠，显著提升代理的决策能力。

DataPrep-Bench: Benchmarking LLMs as Training Data Preparators

👍 49

05/19 08:00

The quality of training data fundamentally determines the capabilities of large language models (LLMs), yet no unified benchmark exists to measure how well LLMs, agents, and data-centric workflows actually prepare training data end to end. We view LLM-driven data preparation as comprising two comple

中文介绍 构建DataPrep-Bench，首个端到端评测LLM作为训练数据准备器的统一基准。涵盖数据清洗、标注、增强等环节，评估LLM代理在数据准备全流程中的表现，填补该领域评测空白。

Skill Self-Play: Pushing the Frontier of LLM Capability with Co-Evolving Skills

👍 33

07/24 08:00

LLM training is shifting from manual design and annotation to interaction-driven self-evolution. However, existing self-evolutionary methods face a fundamental dilemma between task diversity and verification reliability: environment-bound methods obtain precise feedback but confine learning to narro

中文介绍 提出技能自博弈（Skill Self-Play），LLM通过协同演化多样技能来突破能力边界。解决现有自演化方法中任务多样性与验证可靠性的两难，无需环境约束即可获得精确反馈，推动模型自我进化。

Data Pyramid for Embodied Manipulation

👍 29

07/27 08:00

Multimodal foundation models learned to see and to speak by consuming the whole internet. Embodied agents admit no such shortcut, since they require data that couple observations with physical states and actions. These signals can be provided, to varying degrees, by multiple data sources. In this wo

中文介绍 提出数据金字塔（Data Pyramid）用于具身操作。融合多种数据源（真实物理状态、动作、视觉等），以分层方式提供信号，使具身代理能从多源数据中学习，弥补互联网级数据不可用的缺陷。

Molt: A Scalable PyTorch-Native Training Framework for Agentic Reinforcement Learning

👍 27

07/22 08:00

Agentic reinforcement learning research is constant algorithm modification, new estimators, new pipeline stages, new rollout schemes, and in mainstream frameworks each change threads through layers of trainer, distributed backend, and rollout glue: the cost lands on the researcher at every iteration

中文介绍 提出Molt，可扩展的PyTorch原生训练框架，专为智能体强化学习设计。统一算法修改、新评估器、流水线阶段与 rollout 策略，降低研究者每次迭代的开发成本，提升实验效率。

Scaling Native Multimodal Pre-Training From Scratch

👍 20

07/24 08:00

Although large language models (LLMs) exhibit remarkable reasoning capabilities, their reliance on text-only pre-training restricts the perception of the multimodal physical world. Native multimodal pre-training avoids this limitation by training models from scratch on multimodal inputs, thereby ach

中文介绍 从零开始扩展原生多模态预训练，避免LLM文本-only限制。直接在多模态输入上训练模型，保留视觉等感知能力，探讨规模化预训练对多模态理解的影响，实现更贴近物理世界的推理。

The Physics of Multi-Turn Long-Horizon Planning: From Pre-training to Post-training via Single- and Multi-Teacher On-Policy Agentic Distillation

👍 19

07/27 08:00

Multi-turn long-horizon planning is critical for foundation model agents, yet how to fundamentally improve it remains unclear. Existing models are trained on uncontrollable and opaque Internet data, making it difficult to identify how planning ability is acquired, shaped, and integrated. To address

中文介绍 研究多轮长时规划能力的获取机制，提出单/多教师在线智能体蒸馏方法。从预训练到后训练系统分析规划能力的形成，利用可控蒸馏提升基础模型在复杂环境中的多步决策性能。

Agentic Context Management: Solving Agent Memory and Cost by Treating Them as Lifecycle and Architecture Problems

👍 19

07/23 08:00

Production AI agents' failures are less often due to an inability to reason well and more often because they cannot manage what is in their reasoning context: conversation histories, large prompts, large tool definitions, and ballooning tool outputs. Agents drown in their own accumulating history wh

中文介绍 提出智能体上下文管理（Agentic Context Management），将记忆与成本问题视为生命周期和架构问题。通过结构化管理对话历史、工具输出等上下文信息，防止代理因累积历史而失败，提升推理稳定性和效率。

Interactive Training 2: Auditable Control Plane for Live Model Training

👍 16

07/17 08:00

Experiment trackers show how training is progressing, but changing a live run still usually requires trainer-specific code. We present Interactive Training 2, an open-source control plane for steering training through a shared protocol. Training applications declare which settings and actions they e

中文介绍 提出Interactive Training 2，开源控制平面用于实时干预模型训练。通过共享协议声明训练配置与动作，允许用户在训练过程中动态调整超参数、数据集等，实现可审计的实时训练操控。

Multi-Turn On-Policy Distillation with Prefix Replay

👍 11

07/16 08:00

We study on-policy distillation (OPD) for agentic tasks, where an LLM agent interacts with an environment over multiple turns and a student imitates a teacher over these multi-turn interaction histories. Fully online OPD is costly because each update requires fresh student rollouts through the envir

中文介绍 研究多轮在线蒸馏（OPD）在智能体任务中的应用，提出前缀回放（Prefix Replay）策略。复用历史交互片段减少全在线蒸馏的开销，在保持教师质量的同时提升学生多轮决策能力，降低训练成本。

LAMAR: An Open Language-Aware Multilingual Alignment Reranker

👍 9

07/24 08:00

In multilingual retrieval augmented generation, a retriever can retrieve relevant documents written in multiple languages, which are subsequently reranked before answer generation. However, it remains unclear whether existing multilingual rerankers consider document language when ordering semantical

中文介绍 提出LAMAR，开放的多语言对齐重排序器。在检索增强生成中，考虑文档语言对语义排序的影响，显式对齐跨语言文档的排序分数，提升多语言检索质量，特别是对低资源语言更公平。

O-VAD: Industrial Video Anomaly Detection through Object-Centric Tracking and Reasoning

👍 8

07/20 08:00

Industrial Video Anomaly Detection (IVAD) aims to identify anomalous objects and events in an industrial process, which is crucial for modern manufacturing and quality control systems. Existing VLM-based anomaly reasoning methods are capable of detecting open-ended anomalies in general domains. Howe

中文介绍 提出O-VAD，面向工业视频异常检测的物体中心跟踪与推理方法。结合视觉语言模型与物体跟踪，识别开放域异常事件，解决工业场景中多物体、长时序的异常检测需求，提升质检自动化水平。

Multi-Head Latent Control: A Unified Interface for LLM Agent Decision Making

👍 7

07/15 08:00

Large language models are increasingly deployed as agents, but reliable agentic behavior requires more than next-token prediction. At inference time, it is preferred that an agent can decide whether to proceed with its current reasoning, defer to a stronger model, request additional information, inv

中文介绍 提出多头隐式控制（Multi-Head Latent Control），为LLM智能体决策提供统一接口。在推理时允许代理决定继续推理、征求更强模型、请求信息或启动工具调用，通过隐式头实现灵活行为控制。

OpenForgeRL: Train Harness-native Agents in Any Environment

👍 7

07/23 08:00

Modern AI agents rely on elaborate inference harnesses such as Claude Code, Codex, and OpenClaw to drive multi-turn reasoning, tool use, and access to external systems. While powerful, these complex harnesses also make agents hard to train end-to-end with open infrastructure, whose SFT/RL stacks can

中文介绍 提出OpenForgeRL，兼容任意环境的智能体训练框架。集成了Claude Code、OpenClaw等推理装备，支持多轮推理、工具调用，并通过开放SFT/RL栈实现端到端训练，降低智能体训练门槛。

FinanceComplexQA: Benchmarking Agentic Reasoning on Industrial-grade Financial Documents

👍 7

07/21 08:00

Agentic Reasoning has become a transformative force in financial analysis due to its ability to integrate large-scale information and generate reliable and accurate content. However, when handling complex real-world problems, different agents still show significant performance variation. In this wor

中文介绍 构建FinanceComplexQA，基准测试工业级金融文档上的智能体推理能力。涵盖复杂报表、合规文本等，评估不同智能体的信息整合与生成表现，揭示金融场景中现有代理的性能差异与改进方向。

SceneActBench: Can Agents Act on the 3D Scenes They See?

👍 6

07/24 08:00

Vision-language model (VLM) agents increasingly use tools to act on 3D scenes rather than only describe them. Existing 3D benchmarks score textual responses or single-object operations, leaving agent action on complete multi-object 3D scenes under evaluated. We present SceneActBench, a benchmark for

中文介绍 提出SceneActBench，基准测试VLM代理在完整3D场景中的行动能力。要求代理根据视觉场景执行多物体操作（如摆放、组合），超越传统单物体描述任务，评估代理对三维空间的理解与执行能力。

IDEAgent: Agentic Quality-Diversity Search for Research Idea Generation

👍 5

07/24 08:00

Large Language Models (LLMs) have significantly automated the process of scientific discovery over the past few years. However, existing systems share one core limitation: they generate and optimize ideas independently for either Quality or Diversity. This often leads to the generation of ideas in c

中文介绍 提出IDEAgent，智能体式质量多样性搜索用于科研想法生成。结合质量与多样性双优化，避免传统系统单一优化导致的同质化，利用LLM生成具有创新性和覆盖度的研究思路，辅助科学发现。

More On An Internal OpenAI Model Hacking Into HuggingFace

@TheZvi · 39.0K 粉丝 · 219.4K 阅 · 510 赞 · 75 转

07/27 03:14

We now have more details of what happened. Every time we learn more details, it somehow makes things seem worse. The remaining details may have to wait a bit. OpenAI: We recognize there are a lot of

中文介绍 OpenAI 内部模型入侵 HuggingFace 事件再曝新细节，每次了解情况反而更糟糕，更多细节待披露。

22580: From GPT2 to Kimi3, Explained

@waterloo_intern · 10.4K 粉丝 · 215.4K 阅 · 660 赞 · 86 转

07/27 23:22

Twenty-two thousand five hundred and eighty. That’s how many GPT-2 (2019) models fit inside KimiK3 (2026). We scaled up by a factor of 22,580 in seven years. But is it just... scale? In this worklog,

中文介绍 从 GPT-2 到 KimiK3，模型规模 7 年扩大 22,580 倍，博主在 worklog 中探讨这种增长是否只是单纯的 scale。

Opus 5 is a really bad model

@HarukaKunori · 241 粉丝 · 129.0K 阅 · 632 赞 · 44 转

07/27 19:07

After trying out Opus 5 for a few hours today, I honestly think Anthropic's benchmark scores are a complete fraud. Sure, the model might have improved in a few areas, but it has regressed unbelievably

中文介绍 博主实测 Opus 5 后直言其基准分数造假，模型在某些领域有改进，但整体严重倒退，表现糟糕。

Graph Engineering Clearly Explained

@akshay_pachaar · 282.5K 粉丝 · 66.4K 阅 · 515 赞 · 74 转

07/26 02:48

Loop engineering got about six weeks in the spotlight before the timeline moved on. On July 18, Peter Steinberger, the person behind OpenClaw, posted a nine-word question. "Are we still talking loops

中文介绍 循环工程（Loop engineering）在推特上短暂爆火后，博主转向解读图工程（Graph Engineering），引用 OpenClaw 作者 Steinberger 的问题引发思考。

Getting the most out of GPT-5.6: Sol, Terra, and Luna

@cerebras · 64.6K 粉丝 · 48.8K 阅 · 550 赞 · 40 转

07/28 03:44

Authors: @0xSero & Zhenwei Gao (@zhennydez) Your Codex subscription now comes with 3 main models, Sol, Terra, and Luna, each independently trained and served, with reasoning dials. Together, they

中文介绍 博主介绍 Codex 订阅附带的三个独立训练模型 Sol、Terra、Luna，各自配备推理调节旋钮，指导如何充分利用 GPT-5.6。

LFM2.5-Encoders for Fast Long-Context Inference on CPU

Hugging Face Blog · 07/28 23:01

中文介绍 LiquidAI推出LFM2.5-Encoders，专注在CPU上实现快速长上下文推理。

Samsung’s chip workers are jumping ship to rival SK Hynix

MIT Tech Review AI · 07/28 17:18

Lee, an engineer at Samsung’s semiconductor division, clocks out when his shift ends. He used to work longer hours, going the extra mile to excel at his projects. But lately, he’s been coming straight home to work on his job application for the chipmaker’s South Korean rival SK Hynix, sharing tips w

中文介绍 三星半导体部门工程师跳槽至竞争对手SK海力士，员工离职潮加剧，反映行业人才竞争激烈。

[AINews] Much ado about Open Weights

Latent Space · 07/28 14:20

Everyone is writing a lot, but only Kimi K3 shipped today

中文介绍 AI社区热议开源权重，但实际仅有Kimi K3模型发布。

Anthropic on open weights 🔓, Kimi releases K3 weights 🤖, MAI Cyber model 🔐

TLDR AI · 07/28 08:00

中文介绍 Anthropic讨论开源权重，Kimi发布K3模型权重，MAI推出Cyber模型。

OpenAI called the Hugging Face attack unprecedented. But we’ve been here before.

MIT Tech Review AI · 07/28 02:00

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Reading OpenAI’s account last week of how some of its models broke their containment and hacked into the computer systems of Hugging Face, another AI company, was

中文介绍 OpenAI称其模型突破限制入侵Hugging Face系统为「前所未有」，但类似攻击早有先例，引发安全反思。

The path to artificial superintelligence

MIT Tech Review AI · 07/27 20:00

Imagine a healthcare system made up of multiple AI agents: one that manages symptom assessment, another scheduling, a third insurance, and a fourth pharmacy. Each is an expert in its domain. But they all have their own distinct knowledge and objectives. Today they can exchange data, but they are not

中文介绍 文章探讨通往超级人工智能的路径，描述由多个AI代理组成的医疗系统，各自负责不同任务但需协同。

Closing the data loop in AI-driven drug discovery

MIT Tech Review AI · 07/27 19:40

Drug discovery is a high-cost, high-risk endeavor that is under growing pressure from a market increasingly defined by first-mover advantage. Since the 1950s, the cost of developing new pharmaceuticals has roughly doubled every nine years—a phenomenon known as Eroom’s Law. Today, bringing a new drug

中文介绍 AI驱动的药物发现面临数据闭环挑战，药物研发成本每九年翻倍（埃鲁姆定律），需通过数据循环降低成本。

Building the enterprise environment for agentic AI

MIT Tech Review AI · 07/27 19:32

For the enterprise, the promise of agentic AI is much more than just a better chatbot. It is software agents that execute business tasks end-to-end across people, business workflows, data, and systems. The platform best-suited to run agents is built with proper CPU capacity, resilient data access, p

中文介绍 企业级代理AI需构建端到端执行任务的平台，具备充足CPU、弹性和安全环境，以整合人员、流程和数据。

NVIDIA Cosmos-H-Dreams: Bringing Real-Time Generative Simulation to Surgical Robotics

Hugging Face Blog · 07/27 17:32

中文介绍 NVIDIA推出Cosmos-H-Dreams，将实时生成式仿真引入手术机器人领域。

not much happened today

Smol AI News · 07/27 13:44

**Moonshot** released the **Kimi K3** open-weights model, a **2.8T-parameter MoE** with **104B active parameters**, **896 experts**, and **1M-token context** featuring native visual understanding. The release includes open-source infrastructure like **FlashKDA**, **MoonEP**, and **AgentENV**, enabli

中文介绍 Moonshot发布Kimi K3开源权重模型，参数规模2.8T、活跃参数104B、896个专家、支持1M上下文及原生视觉理解，并开源FlashKDA等基础设施。

How AI is expanding what people do at work

OpenAI News · 07/27 11:30

New OpenAI research shows how AI is expanding what workers do, with ChatGPT users taking on tasks across roles and reshaping job boundaries.

中文介绍 OpenAI研究显示AI正扩展员工工作内容，ChatGPT用户跨角色承担任务，重塑工作边界。

Claude Opus 5 🧠, inside OpenAI’s hack 👨‍💻, NVIDIA open weights 🔓

TLDR AI · 07/27 08:00

中文介绍 Claude Opus 5发布，OpenAI黑客事件内幕披露，NVIDIA开源权重。

[AINews] Claude Opus 5: Fable-level performance at Opus price (half Fable)

Latent Space · 07/25 15:25

ain't nobody beats Anthropic at distilling Fable!

中文介绍 Anthropic发布Claude Opus 5，性能达到Fable级别但价格仅为一半，凸显其蒸馏技术优势。

Opus 5

Smol AI News · 07/24 13:44

**Anthropic** launched the **Claude Opus 5** model, which sparked mixed reactions including benchmark scrutiny and praise for its coding-agent capabilities. The model achieved an **Epoch Capabilities Index (ECI) of 159**, slightly below **Fable 5's 161**, but matched Fable 5 on software engineering

中文介绍 Anthropic推出Claude Opus 5，Epoch能力指数159（Fable 5为161），在多项基准上持平，编码代理能力获赞，市场反应不一。

[AINews] Black Forest Labs FLUX 3 - Multimodal Flow Models that beat Seedance 2.0, Gemini Omni and Grok Imagine, and FLUX-mimic video-action robotics model

Latent Space · 07/24 12:30

A HUGE win for BFL!

中文介绍 Black Forest Labs发布FLUX 3多模态流模型，性能超越Seedance 2.0、Gemini Omni及Grok Imagine，并推出机器人动作模仿模型FLUX-mimic。