2026-10-14

Jeva: System One Models from First Principles

👍 266

10/12 20:00

System one models enable direct, low-latency decision-making. Yet recent practice of predicting an answer from serialized text does not, by itself, respect the structure of a decision problem. We argue that current designs underemphasize four important structural requirements. (i) null-aware: Models

中文介绍 针对系统一模型在低延迟决策中的应用，论文指出现有基于序列文本的预测方法未能充分考虑决策问题的结构要求。作者提出了一种新方法，通过增强模型的空值感知能力、模糊性理解和情境推理等特殊结构，从而提高决策效率。本研究的重要意义在于推动智能决策系统的实际应用，特别是在实时反应的场景中。

Agent-Reflex: Teaching Language Agents to Act Reflectively

👍 193

10/12 20:00

Large language models have evolved from single-turn chatbots into interactive agents capable of long-horizon reasoning. In real deployments, MCP tool services can return noisy feedback and users may intervene to correct the agent, but existing methods provide limited supervision for reflection learn

中文介绍 论文提出了一种新方法—Agent-Reflex，通过反思学习训练语言模型，使其更好地应对用户反馈和不确定环境。该方法利用强化学习与交互反馈机制，增强模型的主动适应能力。实验证明，该模型在复杂的决策场景中表现出更高的灵活性和稳定性，促进了语言智能体在实用应用中的发展。

Urbex: Agentic Spatial Grounding for City-Scale 3D Scenes

👍 109

10/12 20:00

City-scale 3D grounding is commonly formulated as candidate ranking, which requires precomputed instance proposals and becomes inefficient in large outdoor scenes. We propose Urbex, an agentic vision-language framework that instead casts grounding as active spatial search. Urbex represents a city-sc

中文介绍 提出Urbex框架，将城市规模的3D场景锚定问题转变为主动空间搜索任务，从而解决了传统候选排名方法的效率低下问题。Urbex结合视觉与语言模型，能在大规模环境中动态适应并高效完成场景理解与定位。此研究推动了增强智能体在城市环境中的应用，为智能导航系统提供了新思路。

Tev: Scaling Long-Horizon Agentic Environments with Dense Progress Rewards

👍 75

10/12 20:00

Long-horizon agents have drawn growing attention for their ability to work autonomously -- from hours to days -- on economically valuable tasks, a capability essential to automated discovery and recursive self-improvement (RSI). Yet open-source models still struggle on long-horizon tasks, largely fo

中文介绍 长时间运行的自主代理在经济价值任务中表现出色，但现有开放源模型在长时间任务中仍面临挑战。论文提出Tev框架，通过稠密进展奖励机制，提升智能体在长期任务中的有效性和自主性。实验证明该方法显著提高了任务完成率，支持了自主发现和自我改善的策略的发展，推动了长期任务处理的研究方向。

EviScene: Beyond "Is It AI?" for Agentic Image Factuality Assessment

👍 51

10/09 20:00

Determining whether an image is AI-generated does not establish whether the situation it depicts is factually supported. We formulate Image Factuality Assessment: given an image and a generic verification instruction, a system returns a supported or refuted verdict with an evidence-grounded explanat

中文介绍 EviScene研究图像事实性评估，超越了简单判别图像是否为AI生成的局限。通过提供一种基于证据的解释机制，该系统能在给定的图像和验证指令下，返回基于事实支持或反驳的判断。此方法在提升AI生成内容可信度的同时，扩展了事实评估的应用范围，推动智能媒介的社会责任研究。

Last Secure Code Benchmark: Is the Code You Vibe Secure?

👍 48

10/09 20:00

AI coding agents now write much of the code that ships, yet whether that code is secure remains unclear. We introduce the Last Secure Code Benchmark, a comprehensive evaluation of secure code generation by coding agents. Built from 150 real vulnerabilities, its 450 tasks span seven widely used langu

中文介绍 引入Last Secure Code Benchmark，为AI代码生成的安全性提供全面评估框架，基于150个真实漏洞和450个任务进行评测。该基准使得开发者能够检测编写代码的安全性，提供了对现有AI编程能力的深入理解，并对未来的安全编程工具开发具有重要意义。

Self-Distillation with Verbal Regularization

👍 29

10/11 20:00

On-policy self-distillation (OPSD) trains a large language model using supervision from a teacher instantiated from the same model as the student. The teacher is conditioned on privileged information (PI), such as verified solutions. The student, which does not observe PI, learns from the teacher's

中文介绍 该研究提出了一种通过在策略学习中引入验证性正则化的方法，提升了基于教师模型指导的自我蒸馏效果。通过有选择地使用特定信息进行训练，模型在保持高效学习的同时，显著增强了推理与决策的准确性，展现出在动态环境中决策支持的潜力。

Data-Centric Recursive Self-Improvement for Navigation via Policy-Synchronized Language-Action Dual Bootstrapping

👍 25

10/10 20:00

Vision-language navigation requires agents to follow natural-language instructions to reach target locations, making high-quality language supervision essential. Since human annotation is costly, prior work has explored scalable data synthesis methods. However, existing methods typically either use

中文介绍 本研究提出了一种数据中心的递归自我改善方法，针对导航任务进行语言-动作的双重引导学习。通过同步策略，从而降低了人类注释的需求，并有效提升了模型理解和执行自然语言指令的能力。这一方法为基于语言的导航任务提供了可扩展的解决方案，彰显了数据驱动AI的发展趋势。

ReTerm: Recovery Training for Self-Improving Terminal Agents

👍 19

10/12 20:00

Terminal agents have shown strong abilities in complex coding tasks. However, they often struggle to recover from errors during long-running execution (e.g., command errors, faulty implementations), eventually leading to task failures. To this end, we present ReTerm, a self-improving terminal agent

中文介绍 ReTerm方法专注于提升终端智能体在长时间执行中的错误恢复能力，避免任务失败。该系统通过自我改善机制，减少由于执行错误导致的性能下降。研究显示，此方法在复杂编程任务中的有效性和鲁棒性，为终端智能体的发展提供了新路径。

Open-o3-Video-v2: Seek, Ground, and Reason with Verifiable Spatio-Temporal Evidence

👍 18

10/12 20:00

Reliable video reasoning requires models to produce correct answers and expose supporting visual evidence. Open-o3-Video enables verifiable reasoning by embedding timestamps and object-level bounding boxes into reasoning traces, but is restricted to frames from an initial sparse sampler. We introduc

中文介绍 Open-o3-Video提出了一种可验证视频推理的新框架，通过嵌入时间戳和目标级边界框，实现了对推理过程的可追溯性。该模型在高效利用时序信息的同时，增强了对生成答案的可信度，推动了视频智能分析领域的研究进步。

LightVAE: Towards Compact and Efficient Video Autoencoders

👍 16

10/09 20:00

Latent diffusion models enable high-quality video generation, but video VAE encoding and decoding remain computationally expensive. Existing methods use stage-specific lightweight operators, yet either constrain spatial and temporal computation jointly or remove temporal computation altogether, with

中文介绍 LightVAE提出了一种高效的变分自编码器，旨在降低视频生成的计算开销。通过优化潜在扩散模型，减少了空间和时间计算的耦合问题，使得视频编码和解码过程更加高效。此研究为可扩展的视频生成技术奠定了基础，影响了视频内容创作的未来发展。

LifeSearch: Benchmarking Deep Research Agents for Real-World Life and Productivity

👍 15

10/12 20:00

Deep research (DR) agents address open-ended information needs through multi-step reasoning, long-horizon web search, and evidence synthesis. However, existing benchmarks primarily focus on academic research or synthetic tasks, leaving DR agents' capabilities in real-world life and professional prod

中文介绍 LifeSearch提出了一个针对深度研究代理的基准，专注于真实生活和生产力的多步骤推理和网络搜索。通过结合实际应用案例，评估了深度研究代理在现实世界中的表现，促进了智能代理在非学术领域的应用发展。

LongR: Unleashing Long-Context Reasoning via Reinforcement Learning with Dense Utility Rewards

👍 13

10/11 20:00

Reinforcement Learning (RL) has emerged as a key driver for LLM reasoning. This capability is equally pivotal in long-context scenarios---such as long-dialogue understanding and structured data analysis---where the challenge extends beyond consuming tokens to performing rigorous deduction. While exi

中文介绍 LongR研究了通过强化学习提高长上下文推理效果的方法，专注于长对话理解和结构化数据分析的挑战。该方法将严格推理引入到长时间序列处理之中，显著提高了模型的理解能力，并为复杂长文本处理的研究提供了新的思路。

Pristine Models

👍 12

10/08 13:58

Training on broad, unfiltered data enables high-quality image generation, but can also teach models concepts that should not be generated; restricting the training data ensures safety, but reduces quality because examples containing unsafe concepts also provide useful knowledge that is not itself un

中文介绍 Pristine Models强调在广泛未过滤数据上训练以提高图像生成质量，同时平衡安全性和输出质量。在确保不再生成不当内容的同时，保留有用知识，展示了在数据选择与安全性之间的权衡对AI生成模型的影响。

Latent World Action Model for GUI Agents

👍 12

10/08 20:00

Pixel-space world models have recently emerged as a promising approach for GUI agents, predicting future screens prior to execution and subsequently re-perceiving them for action selection. However, this two-stage paradigm faces two key challenges in GUI environments: (L1) the high computational cos

中文介绍 隐潜世界行动模型被提出用于GUI智能体，旨在通过预测未来界面并重新感知以选择动作。然而，该模型在计算复杂性和环境适应性方面面临挑战。此研究为GUI环境中的智能交互提供了新的建模思路，有望推动人机交互领域的发展。

AUGUstus: Autonomous Driving VLA Model with Unified Generation and Understanding

👍 10

10/11 20:00

Vision-language-action (VLA) models show strong potential for end-to-end autonomous driving by leveraging world knowledge and understanding to excel in long-tail scenarios. However, current VLAs have two limitations: 1) Degeneration into vision-action (VA) models. For low latency, regression/flow-ba

中文介绍 AUGUstus提出了一种新型的视觉-语言-动作(VLA)模型，用于实现自动驾驶。该模型结合环境知识，克服了低延迟场景中的退化问题，展现了在复杂驾驶环境中的优秀表现。研究推动了自动驾驶技术的发展，并强调了VLA模型在现实应用中的重要性。

What Really Works for Universal Multimodal Embeddings? A Controlled Study of Contrastive Training Strategies

👍 10

10/09 20:00

Contrastive learning has become the standard paradigm for multimodal representation learning, aligning semantically matched pairs while separating mismatched ones in a shared embedding space. Although various contrastive training strategies have been proposed across negative selection, batch constru

中文介绍 本研究对对比学习在多模态表示学习中的有效策略进行了受控研究，探讨了匹配对的语义对齐和不匹配对的分离方法。研究结果表明，不同的对比训练策略在任务表现中存在显著差异，为多模态表示学习提供了新参考，推动了表示学习方法的发展。

EMHO: EMbodied Agent Harness Optimization via Experience Traces

👍 10

10/05 20:00

Improving embodied agents often focuses on optimizing the underlying model through training, while the surrounding agent harness that controls planning, context, and tool use is typically engineered. We ask whether this harness can instead improve itself directly from experience traces under sparse

中文介绍 EMHO研究了体态代理的优化问题，聚焦于如何通过经验回溯直接改善代理的工作表现。研究成果表明，代理能够通过经验学习优化计划、上下文和工具使用的能力，为提升体态智能体的有效性提供了全新思路。

WorldRecon: Generative Multi-View Mesh Reconstruction for Large-Scale 3D World

👍 10

10/12 20:00

Reconstructing complete, high-quality meshes of large-scale scenes from unposed multi-view images requires faithful geometric grounding while plausibly completing surfaces that are unobserved or only sparsely captured. We present WorldRecon, a generative multi-view reconstruction framework for recov

中文介绍 WorldRecon提出了一种生成性多视角网格重建框架，用于从多视角图像重建大规模场景的高质量网格。该方法通过确保几何基础的真实性，实现了对未观察到的区域表面合理填补，推动了3D场景重建技术的进步，具有广泛的应用前景。

SWE-MISBEHAVIOR: Benchmarking and Mitigating Boundary Violations in Software Coding Agents

👍 9

10/12 20:00

We present SWE-MISBEHAVIOR, the first taxonomy and evaluation framework for diagnosing boundary-violating misbehaviors in coding agents. It contains 7.4K human-in-the-loop synthesized samples spanning three layers of agent action space: (i) Code Generation, (ii) File Management, and (iii) System Int

中文介绍 SWE-MISBEHAVIOR构建了第一个用于诊断编码代理中的边界违反行为的分类体系和评估框架，包含7.4K人机协同合成样本，覆盖了代码生成、文件管理及系统交互等任务层面。该研究为开发更可靠的编码智能体奠定了基础，具有重要的实践意义。

How Grok Bot Changed the Way I Work

@melvindvivas · 6.3K 粉丝 · 209.9K 阅 · 541 赞 · 83 转

10/13 01:20

I've been using AI coding assistants like Claude Code, Codex, and Cursor for quite some time. They've already changed how I build software, but Grok Bot has changed something else: how I work

中文介绍 博主分享了他使用 Grok Bot 代替传统 AI 编程助手（如 Claude、Codex 和 Cursor）后改进的工作方式，强调了这种工具在软件开发流程上的革命性影响。

Introducing Glass: Learn Things Once.

@eglyman · 393.9K 粉丝 · 88.6K 阅 · 724 赞 · 58 转

10/13 10:00

Last week, David Senra asked me: “How does Ramp move so fast?” Today we’re launching the answer. But before we do, I’d like to tell you why we built it. Giving every employee AI — the best models, the

中文介绍 博主介绍了新推出的 Glass 工具，该工具旨在提升员工在工作中的效率，充分利用 AI 模型来加速企业运作。

35+ UNCENSORED AI SITES FOR 18+ IMAGE AND VIDEO GENERATION

@rexsooo · 8.5K 粉丝 · 83.4K 阅 · 703 赞 · 46 转

10/12 12:52

Adult AI image and video generators, NSFW character chat, anime tools, and AI companions This list covers 35 uncensored or NSFW-capable AI platforms across image generation, video creation, anime, AI

中文介绍 这是一份关于 35 个成人 AI 图像和视频生成平台的汇总，涵盖不限于 NSFW 内容的多种生成工具，反映了行业中对此类平台的需求和关注。

How to make a good RL task

@GottliebEli · 809 粉丝 · 28.9K 阅 · 506 赞 · 37 转

10/13 14:36

I’ve been making lots of tasks lately. Many poor, some decent, and now finally, a few good. Through this work I’ve gained what I believe to be valuable insights into the task generation process that

中文介绍 博主分享了在生成强化学习任务中的探索经历，提供了一些实用的见解，特别是在创建高质量任务方面的关键要点。

Gemini 4 Argon available on Google Cloud, built to help solve your business’s hardest problems

@googlecloud · 573.7K 粉丝 · 21.2K 阅 · 502 赞 · 70 转

10/13 12:07

Last week we announced that Gemini Enterprise has evolved to make it even simpler to get work done, and introduced the Gemini agent, your new universal agent for work. But model intelligence underpins

中文介绍 谷歌云推出了 Gemini 4 Argon，旨在帮助企业解决复杂问题，并强调新引入的通用代理可以简化工作流程，此举标志着 AI 在企业中的应用愈发普及。

Gemini 4 Argon: Built to solve your hardest problems

@GoogleAIStudio · 208.3K 粉丝 · 19.6K 阅 · 635 赞 · 45 转

10/13 13:33

Gemini 4 Argon is now globally available today for developers and enterprises via the API in AI Studio and Gemini platform, for Google AI Ultra subscribers in Google Antigravity, and Google AI Ultra

中文介绍 Google AI Studio 发布 Gemini 4 Argon，面向开发者和企业开放，展现了其在 AI 领域的前沿技术和强大支持，表明智能模型在商业应用中的重要性。

How to Build an AI Hacking Agent That Finds Bugs While You Sleep

@zack0x01_ · 5.3K 粉丝 · 19.3K 阅 · 501 赞 · 71 转

10/13 02:52

I built an AI agent that hunts bugs while I sleep. Not a chatbot that gives me advice about hacking, an agent on my own server with hands. It maps a target, tests the bug classes I actually care

中文介绍 博主分享了自己构建的 AI 漏洞猎人，该智能代理可以在他睡觉时自动寻找漏洞，展示了 AI 在网络安全领域的创新应用潜力。

Crossing the Threshold: How OpenAI and Anthropic Approach Autonomy and Safety in Financial Agents

Latent Space · 10/14 17:40

At AI Engineer New York, frontier labs and financial firms explored how to give agents useful autonomy while keeping consequential decisions under control.

中文介绍 在纽约的AI工程师会议上，前沿实验室和金融公司讨论了如何在保持重要决策控制的同时，赋予代理人有效的自主性。

Lab-in-the-Loop at Compute Speed — Ian Quigley, Leash Bio

Latent Space · 10/14 16:35

How Leash Bio tamed DELs, the assay that traumatizes everyone, and used the data to catch cheating models. And why that same data is perfect for AI agents learning to think like chemists.

中文介绍 Leash Bio成功解决了数据采集中的难题，利用该数据帮助AI代理学习化学思维。

HiRO-ACEv2: Simulating extreme weather at 3 km resolution for a fraction of the cost

Hugging Face Blog · 10/14 12:12

中文介绍 HiRO-ACEv2能够以低成本在3公里分辨率下模拟极端天气。

Amp puts AI agents to work beyond code with OpenAI models

OpenAI News · 10/14 08:00

See how Amp pushes the frontier of what agents can do with GPT-6 Astra, Ultrafast, GPT-Live-1, and Sign in with ChatGPT.

中文介绍 Amp利用OpenAI的模型，拓展了AI代理的工作范围，超越了代码的界限。

SentinelOne cuts malware analysis from weeks to hours

OpenAI News · 10/14 08:00

With OpenAI Daybreak, SentinelOne cuts weeks of malware analysis to hours, creating capacity to investigate more threats.

中文介绍 SentinelOne通过OpenAI的Daybreak技术，将恶意软件分析时间从数周缩短至数小时，提升了威胁调查能力。

How Third Moment Research runs 10,000 agents for 36+ hours

OpenAI News · 10/14 03:00

With GPT-6 Astra and GPT-6.1 Sol, Codex, and Daybreak for cybersecurity, this AI lab builds agents that focus for days and improve with use.

中文介绍 Third Moment Research使用GPT-6 Astra和其他技术，运行10000个代理，持续工作超过36小时，并通过使用不断提升性能。

How Block is building AI-powered security with OpenAI technology

OpenAI News · 10/14 03:00

Block is building AI-powered security with OpenAI. Its security team investigates threats faster and proposes patches in minutes.

中文介绍 Block正在利用OpenAI技术构建AI安全系统，其安全团队能够在数分钟内调查威胁并提出修复补丁。

[AINews] AI Engineer NYC: Gemini 4 Argon, Vals AI Benchmarks

Latent Space · 10/14 01:40

Gemini 4 Argon went GA today, but we already gave it title story 2 weeks ago and most benchmarks, like CursorBench, simply independently confirm their claims.

中文介绍 Gemini 4 Argon于今日发布，之前的标题故事及独立基准测试已确认其声称的性能。

Google DeepMind and Singapore Government collaborate to build fine-tuned AI model for scam detection

DeepMind Blog · 10/13 22:00

Learn how Google DeepMind and GovTech Singapore uses lightweight on-device Gemma models to intercept conversational scams in private chats with zero cloud data egress.

中文介绍 Google DeepMind与新加坡GovTech合作，使用轻量级Gemma模型在私密聊天中拦截诈骗信息，且不涉及云数据。

The EU’s next chapter in AI is agentic

OpenAI News · 10/13 20:00

New analysis shows how agentic AI adoption is growing across EU enterprises and small businesses—and where deeper use could unlock more value.

中文介绍 新的分析显示，EU企业和小型企业对代理AI的采用正在增长，更深层次的应用可能会释放更多价值。

OpenAI dev arrested 🚨, Pi API 🤖, OpenCode 2.0 🧑‍💻

TLDR AI · 10/13 20:00

中文介绍 OpenAI开发者被捕，Pi API发布，OpenCode 2.0启动。

Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics

Latent Space · 10/13 18:25

Our first robotics pod with Dyna’s Chief Scientist on their big bet on World Action Models & the Next Frontier of Physical AI

中文介绍 Dyna Robotics的首席科学家介绍了其Dyna-2机器人项目，旨在开发具有小时级自主性的多功能物理代理。

Developing Gemini 4 Argon responsibly and safely

DeepMind Blog · 10/13 15:43

中文介绍 DeepMind正在负责和安全地开发Gemini 4 Argon模型。

Run fast, private agents on-device with LFM2.5-24B-A2B

Hugging Face Blog · 10/13 13:05

中文介绍 LFM2.5-24B-A2B使得在设备上运行快速、私有的代理成为可能。

Gemini 4 Argon for Robotics

DeepMind Blog · 10/13 12:01

中文介绍 Gemini 4 Argon被应用于机器人技术，推动相关研究与开发。