2026-06-01

AgentDoG 1.5: A Lightweight and Scalable Alignment Framework for AI Agent Safety and Security

👍 127

05/28 08:00

Modern open-world agents such as OpenClaw exhibit powerful cross-environment execution capabilities yet introduce broad new safety risk sources. Meanwhile, advanced frontier AI models drastically lower attack barriers, rendering current agent alignment frameworks inadequate for real-world deployment

中文介绍 AgentDoG 1.5 提出轻量可扩展对齐框架，应对开放世界智能体（如 OpenClaw）引入的安全风险，保障实际部署中的智能体安全。

OmniRetrieval: Unified Retrieval across Heterogeneous Knowledge Sources

👍 66

05/28 08:00

Real-world information needs require access to structurally diverse knowledge sources, from unstructured text and relational tables to knowledge graphs and property graphs. Existing retrievers, however, operate over one source at a time under a fixed query language, leaving the broader landscape of

中文介绍 OmniRetrieval 实现跨异构知识源（非结构化文本、关系表、知识图谱、属性图）的统一检索，突破传统检索器单一源与固定查询语言的限制。

YoCausal: How Far is Video Generation from World Model? A Causality Perspective

👍 41

05/28 08:00

As video diffusion models (VDMs) advance toward world models, a key question arises: do they truly understand causality, or merely overfit to statistical temporal patterns? Existing benchmarks mostly rely on synthetic data, limiting real-world generalization due to the sim-to-real gap. We present Yo

中文介绍 YoCausal 从因果视角评估视频扩散模型的世界模型能力，揭示其是否真正理解因果关系还是过拟合时间统计模式，并克服合成数据评估的局限。

Why Far Looks Up: Probing Spatial Representation in Vision-Language Models

👍 38

05/28 08:00

Vision-language models (VLMs) achieve strong performance on spatial reasoning benchmarks, yet it remains unclear whether this reflects structured 3D understanding or reliance on statistical shortcuts in natural images. We introduce a representation-level analysis framework that constructs minimal co

中文介绍 提出表征级分析框架，通过最小控制实验探究视觉语言模型的空间表征，区分结构化 3D 理解与图像统计捷径依赖。

GenClaw: Code-Driven Agentic Image Generation

👍 31

05/28 08:00

Image generation models have evolved from text-conditioned pixel synthesis toward multimodal agents endowed with visual comprehension and tool invocation capabilities. Yet, existing agents remain at the mercy of underlying black-box image models. Their workflow is trapped in a repetitive cycle of pr

中文介绍 GenClaw 通过代码驱动智能体图像生成，摆脱黑盒模型的重复提示循环，实现更可控、可编辑的生成流程。

EarlyTom: Early Token Compression Completes Fast Video Understanding

👍 27

05/28 08:00

Video large language models (Video-LLMs) have demonstrated strong capabilities in video understanding tasks. However, their practical deployment is still hindered by the inefficiency introduced by processing massive amounts of visual tokens. Although recent approaches achieve extremely low token ret

中文介绍 EarlyTom 采用早期视觉 token 压缩策略，大幅降低 Video-LLM 处理海量 token 的计算开销，实现快速视频理解。

UniSteer: Text-Guided Flow Matching in Activation Space for Versatile LLM Steering

👍 21

05/28 08:00

Activation-based control steers large language models (LLMs) by intervening on their internal representations during inference, and has emerged as an effective paradigm for controlling behaviors such as persona and style. However, existing methods often rely on fixed steering directions or task-spec

中文介绍 UniSteer 用文本引导的激活空间流匹配，实现 LLM 多样化行为控制（如人格、风格），克服固定方向或任务特定方法的局限性。

Skill0.5: Joint Skill Internalization and Utilization for Out-of-Distribution Generalization in Agentic Reinforcement Learning

👍 21

05/27 08:00

Equipping large language models with explicit skills has emerged as a promising paradigm for enabling autonomous agents to solve complex tasks. Agent skills can be inherently divided into general skills for broad cognitive transfer and task-specific skills for dynamic execution. However, existing sk

中文介绍 Skill0.5 联合内化与利用通用和任务特定技能，提升智能体强化学习在分布外场景下的泛化能力。

LoMo: Local Modality Substitution for Deeper Vision-Language Fusion

👍 20

05/28 08:00

Vision-Language Models (VLMs) have achieved substantial progress across a wide range of understanding and reasoning tasks, driven by large-scale image-text training aimed at multimodal fusion. Ideally, replacing a textual question with its rendered-image counterpart should leave model performance es

中文介绍 LoMo 通过局部模态替换（将文本问题替换为渲染图像）测试视觉语言模型的多模态融合深度，揭示模型对文本捷径的依赖。

Colored Noise Diffusion Sampling

👍 18

05/28 08:00

Diffusion models achieve state-of-the-art image synthesis, with their generative trajectories fundamentally exhibiting a spectral bias, resolving low-frequency global structures early and high-frequency fine details later. Conventional stochastic differential equation (SDE) solvers fail to account f

中文介绍 Colored Noise Diffusion Sampling 利用扩散模型的频谱偏差（先低频后高频），改进传统 SDE 求解器，提升图像生成质量。

Xetrieval: Mechanistically Explaining Dense Retrieval

👍 17

05/28 08:00

Explaining why dense retrievers assign high relevance scores remains challenging because retrieval decisions are made through opaque high-dimensional embeddings. Existing explanations often focus on surface signals, such as lexical matches, token alignments, or post-hoc textual rationales, and thus

中文介绍 Xetrieval 从机制层面解释稠密检索模型的关联评分，揭示其决策依据，超越词汇匹配、token 对齐等表面信号。

CausaLab: A Scalable Environment for Interactive Causal Discovery Toward AI Scientists

👍 13

05/28 08:00

We introduce CausaLab, a scalable environment for evaluating interactive causal discovery by LLM agents. Unlike prior evaluations, CausaLab evaluates both whether an agent can solve a problem using causal evidence and whether its answer is grounded in a faithful recovered causal mechanism. Each epis

中文介绍 CausaLab 提供可扩展的交互式因果发现环境，评估 LLM 智能体解决因果问题的能力及答案的因果忠实性。

When Cloud Agents Meet Device Agents: Lessons from Hybrid Multi-Agent Systems

👍 11

05/28 08:00

The design space of agentic AI inference spans two extremes: frontier large language models (LLMs), typically hosted in the cloud and offering strong performance across a wide range of tasks at substantially high cost, and more cost-efficient small language models (SLMs), which are amenable to on-de

中文介绍 探讨云端强模型与设备端轻量模型混合的多智能体系统设计，平衡性能与成本，为实际部署提供经验。

PRISM: A Multi-Dimensional Benchmark for Evaluating LLM Peer Reviewers

👍 10

05/27 08:00

The rapid growth in submissions to machine learning venues has strained the scientific peer-review system and intensified interest in LLM-based automated peer reviewers. However, how good these systems are actually, especially compared to human reviewers at catching scientific gaps, remains poorly u

中文介绍 PRISM 多维基准评估 LLM 同行评审质量，与人类评审者在发现科学漏洞、论证完整性等维度进行对比。

Towards Verifiable Multimodal Deep Research: A Multi-Agent Harness for Interleaved Report Generation

👍 10

05/28 08:00

Large Language Models (LLMs) have advanced autonomous agents from deep search, which retrieves concise factual answers, to deep research, which synthesizes scattered evidence into long-form reports. However, verifiable multimodal deep research remains challenging due to open-ended synthesis without

中文介绍 提出可验证多模态深度研究的多智能体框架，从深度搜索走向长文报告生成，支持证据交叉验证与模态交错。

UI-KOBE: Knowledge-Oriented Behavior Exploration for Lightweight Graph-Guided GUI Agents

👍 10

05/28 08:00

Recent advances in mobile GUI agents have shown strong potential for automating mobile tasks, but most effective systems still depend on large vision-language models for screenshot understanding and long-horizon planning. Small GUI agents that can be deployed directly on mobile devices are more attr

中文介绍 UI-KOBE 通过知识导向行为探索和图引导，实现轻量级移动 GUI 智能体，无需大视觉模型即可自动完成任务。

RUBRIC-ARROW: Alternating Pointwise Rubric Reward Modeling for LLM Post-training in Non-verifiable Domains

👍 9

05/27 08:00

Pointwise reward modeling offers critical signals for LLM post-training, yet struggles with absolute scoring in subjective, non-verifiable settings. Rubric-based methods address this by decomposing evaluation into explicit criteria, but existing approaches typically depend on frontier LLMs and suffe

中文介绍 RUBRIC-ARROW 通过交替点式评分标准奖励建模，提升 LLM 在主观非可验证领域的后训练效果，降低对前沿模型的依赖。

PhyGenHOI: Physically-Aware 4D Generation of Dynamic Human-Object Interactions

👍 8

05/28 08:00

We address the task of generating physically accurate and visually faithful 4D Human-Object Interaction (HOI). Given a static 3D human and target object represented as 3D Gaussian Splats (3DGS), our goal is to synthesize dynamic scenes where the human actively engages with the object through actions

中文介绍 PhyGenHOI 基于 3D 高斯泼溅实现物理感知的 4D 人-物交互生成，从静态人体和目标物体合成动态动作。

DynaFLIP: Rethinking Robotics Perception via Tri-Modal-Dynamics Guided Representation

👍 7

05/28 08:00

Robot manipulation critically depends on perception that preserves the action-relevant aspects of a scene. Yet most robot learning pipelines are built upon visual encoders pre-trained for static recognition or vision-language alignment, leaving motion understanding to downstream policies. We introdu

中文介绍 DynaFLIP 通过三模态动力学引导表示学习，增强机器人操作感知对动作相关信息的保留，超越静态识别预训练。

WorldMemArena: Evaluating Multimodal Agent Memory Through Action-World Interaction

👍 7

05/28 08:00

Multimodal large language models are increasingly deployed as long-horizon agents, where memory must do more than recall: it must track an evolving world, revise what has gone stale, and surface the right evidence at decision time. Existing benchmarks measure recall over static dialogue, collapse me

中文介绍 WorldMemArena 通过动作-世界交互评估多模态智能体记忆的动态追踪、更新和决策证据召回能力。

[AINews] Founders and Forward Deployed Engineers

Latent Space · 05/30 09:57

a quiet day lets us highlight the new AIE WF focuses

中文介绍 当日市场平静，新闻聚焦AIE（人工智能工程）工作流的新重点。

Boston Children’s uses AI to unlock new diagnoses

OpenAI News · 05/29 20:00

Boston Children’s Hospital uses OpenAI technology to improve patient care, reduce operational burden, and help diagnose more than 40 rare disease cases.

中文介绍 波士顿儿童医院利用OpenAI技术改善患者护理、降低运营负担，并帮助诊断超过40例罕见疾病。

How Braintrust turns customer requests into code with Codex

OpenAI News · 05/29 20:00

How Braintrust engineers use Codex with GPT-5.5 to run experiments and code faster.

中文介绍 Braintrust工程师使用Codex与GPT-5.5，将客户请求转化为代码，加速实验和编码过程。

How the Pope’s Magnifica Humanitas offers a template for individuals to meet the AI moment

MIT Tech Review AI · 05/29 18:00

Pope Leo XIV’s new encyclical on artificial intelligence includes a statement that warrants serious attention from technologists and policymakers: “Technology is never neutral.” Magnifica Humanitas (“Magnificent Humanity”) is a clarion call to all people to act with courage and solidarity as we ente

中文介绍 教皇利奥十四世发布关于人工智能的通谕《尊贵人性》，指出「技术绝非中立」，呼吁所有人勇敢应对AI时代。

not much happened today

Smol AI News · 05/29 13:44

**Anthropic** rolled out **Claude Opus 4.8**, which shows incremental improvements but mixed benchmark results, including better cooperation and coding behavior but some regressions in document parsing. Platform updates include mid-conversation system instructions enhancing long agent sessions, thou

中文介绍 Anthropic推出Claude Opus 4.8，带来渐进改进，合作与编码行为提升，但文档解析出现回退。平台更新包括对话中系统指令增强。

Strengthening societal resilience with Rosalind Biodefense

OpenAI News · 05/29 11:00

OpenAI launches Rosalind Biodefense, expanding trusted access to GPT-Rosalind for vetted developers and U.S. government partners advancing biodefense, public health, and pandemic preparedness through frontier AI.

中文介绍 OpenAI推出Rosalind Biodefense，向经过审查的开发者和美国政府合作伙伴扩展GPT-Rosalind的信任访问，以推进生物防御、公共卫生和大流行病防范。

[AINews] Anthropic raises $965B Series H, releases Opus 4.8 and Dynamic Workflows/ultracode

Latent Space · 05/29 10:07

Total Anthropic victory!

中文介绍 Anthropic完成9650亿美元H轮融资，并发布Opus 4.8和Dynamic Workflows/Ultracode。

A shared playbook for trustworthy third party evaluations

OpenAI News · 05/29 08:00

OpenAI shares guidance on third-party AI evaluations, covering how to assess model capabilities, safeguards, and validity for frontier systems.

中文介绍 OpenAI发布关于第三方AI评估的指南，涵盖如何评估前沿系统的模型能力、安全防护和有效性。

Profiling in PyTorch (Part 1): A Beginner's Guide to torch.profiler

Hugging Face Blog · 05/29 08:00

中文介绍 PyTorch profiling入门指南，介绍torch.profiler的基本用法。

Opus 4.8 🧠, Anthropic at $965B 💰, Microsoft’s coding model 👨‍💻

TLDR AI · 05/29 08:00

中文介绍 当日AI新闻汇总：Anthropic Opus 4.8发布，Anthropic估值9650亿美元，微软推出编程模型。

The Age of Async Agents — Cognition's Walden Yan & OpenInspect's Cole Murray

Latent Space · 05/29 02:41

80% Devin Commits, Spec-to-PR Workflows, Full VMs, Agent Memory, and PMs Shipping Code

中文介绍 Cognition和OpenInspect探讨异步代理时代，涉及80% Devin提交、Spec-to-PR工作流、完整虚拟机、代理内存和PM交付代码。

How Endava builds an agentic organization with Codex

OpenAI News · 05/28 20:00

Learn how Endava uses Codex to build an agentic organization, accelerating software delivery and reducing requirements analysis from weeks to hours.

中文介绍 Endava利用Codex构建代理型组织，将软件开发交付加速，需求分析从数周缩短至数小时。

The AI Hype Index: AI gets booed in graduation season

MIT Tech Review AI · 05/28 17:51

It is one thing to say AI will change the world. It is another to expect the class of 2026 to applaud it. In fact, when former Google CEO Eric Schmidt told University of Arizona graduates that their task is to help shape AI, he was met with a resounding chorus of boos. “I can…

中文介绍 AI炒作指数：在毕业季，当谷歌前CEO埃里克·施密特对亚利桑那大学毕业生说他们的任务是帮助塑造AI时，遭到一片嘘声。2026届毕业生对AI持怀疑态度。

[AINews] Cognition raises $1B in $26B Series D

Latent Space · 05/28 15:26

coding is an uncapped TAM market

中文介绍 Cognition完成260亿美元D轮融资（其中10亿美元），表示编程市场总规模无上限。

Anthropic raises $65B in Series H at a $965B post-money valuation, releases Opus 4.8 and Dynamic Workflows

Smol AI News · 05/28 13:44

**Anthropic** announced a massive **$65B Series H financing** at a **$965B valuation**, led by **Altimeter, Dragoneer, Greenoaks, and Sequoia**, with run-rate revenue surpassing **$47B**. They launched **Claude Opus 4.8**, an update to Opus 4.7 featuring "sharper judgment," "more honesty," and longe

中文介绍 Anthropic完成650亿美元H轮融资，投后估值9650亿美元，领投方包括Altimeter、Dragoneer、Greenoaks和Sequoia，年化营收超470亿美元。同时发布Claude Opus 4.8，具备更敏锐判断力。