2026-09-27

Training Object Permanence in World Models

👍 202

09/22 20:00

Object permanence and solidity are hallmarks of human cognitive priors. Recent studies show that video generation models, a paradigmatic class of current world models, have begun to show emerged reasoning abilities, making them ideal candidates for building human-like physical intelligence. Do video

中文介绍 论文旨在研究物体的永恒性和实体性，探索视频生成模型在这些方面的推理能力。通过结合行为智能与生成模型，提出了一种新的学习框架，该框架能够通过动态视频生成实现物体持久性的理解。该方法在复杂场景中的应用表明，模型在理解物体状态变更方面取得了显著进展，推动对人类类物理智能的建模研究。意义：为Agent与Reasoning方向提供新的研究视角。

SpeakerMem-R1: Speaker-Centered Dual-Track Memory for Multi-Party Dialogue

👍 84

09/21 20:00

Long-term conversational memory in multi-party settings requires more than retrieving relevant content from long-term conversations: it must distinguish who said what, whom each statement concerns, how individuals perceive one another, what information is shared by the group, and how states change o

中文介绍 提出SpeakerMem-R1，一种面向多方对话的双轨记忆模型，能够有效管理长期对话中的信息。该模型能区分发言者及其观点，处理个体间的感知及信息共享状态的变化，更准确地捕捉对话上下文。实验表明，该模型在多方对话的记忆检索和管理上表现更优，增强了对话系统的交互性和智能性。意义：对Multi-Party Dialogue的研究提供了新的思路。

Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs

👍 75

09/23 20:00

While Large Language Models (LLMs) rely on highly non-linear components, in this work we demonstrate that they exhibit fundamental linearity: when inputs from distinct text streams are linearly combined, the model outputs a superposition of the individual next-token distributions. We term this the S

中文介绍 研究表明大型语言模型（LLMs）在处理非线性输入时，实际上展现出线性叠加的特征。通过对来自不同文本流的输入进行线性组合，模型能够输出各自的下一个token分布的叠加。这一发现扩展了我们对LLMs的理解，为进一步优化模型训练和输出提供了基础，可能在语言生成和理解方面产生重要影响。意义：为LLMs的推理能力研究提供了新的理论基础。

Spatial-Interactor: Learning Spatial Reasoning through Interaction with the Observable Physical World

👍 48

09/18 20:00

Spatial reasoning is essential for vision-language models (VLMs) to understand and act in the physical world. Reasoning in dynamic environments requires VLMs to perceive local state transitions caused by object motion and viewpoint changes and integrate them over long trajectories to maintain an upd

中文介绍 提出Spatial-Interactor，一种通过与可观察的物理世界互动来学习空间推理的模型。该模型能够感知动态环境中的状态变化，并整合长期轨迹信息以进行推理。这种方法提升了视觉语言模型（VLMs）在复杂场景中的表现，促进了对物体运动和视点变化的理解。实验结果表明，模型在各种环境中具有强大的适应能力。意义：对Agent与World Model领域的研究具有重要推动作用。

HappyWorld-Bench

👍 43

09/20 20:00

Evaluating world models requires assessing both the quality of the worlds they generate and their consistency and responsiveness under exploration, interaction, and modification. We introduce HappyWorld-Bench, a comprehensive benchmark that evaluates whether generated worlds remain reliable as agent

中文介绍 为评估世界模型的生成质量及其在探索和交互中的一致性与响应能力，提出了HappyWorld-Bench基准。该基准通过多项指标衡量生成世界的可靠性，帮助研究人员了解模型在复杂任务下的表现。实验结果显示，通过HappyWorld-Bench基准测试，可以更好地评估和优化生成模型，推动智能体对环境的理解与适应。意义：为世界模型的评估与优化提供了重要工具。

The Past Frames the Future: Memory for Autoregressive Video Generation

👍 39

09/22 20:00

Advances in generative models have improved video fidelity, enabling long-horizon generation, interactive world modeling, and evolving visual environments. Autoregressive (AR) video generation extends visual sequences through causal rollouts. However, a fundamental bottleneck emerges: as the generat

中文介绍 针对autoregressive (AR) 视频生成中的长时间生成瓶颈，提出一种新策略，允许模型在生成过程中利用过去的帧信息以增强生成效果。通过对生成过程进行改进，模型在生成的连贯性和质量上显著提升，使得在复杂交互场景中的应用变得更加可行。这项研究为推动视频生成技术及其在动态环境中的应用提供了契机。意义：在视频生成与理解的研究中引领了新方向。

Just-in-Time Memory: Learning to Curate Task-Adaptive Memory for LLM Agents

👍 35

09/22 20:00

Agentic memory systems reuse past experience to improve future performance, yet most existing designs curate memory at write time: once a task is completed, its trajectory is distilled into a fixed artifact, such as a reflection, workflow, skill, or reasoning strategy, that is later retrieved by sim

中文介绍 提出了Just-in-Time Memory系统，通过学习任务适应性记忆以增强LLM代理的性能。该系统在任务完成后动态地管理和储存记忆，能够更灵活地利用历史经验，提高未来任务的解决能力。实验结果表明，模型在多种任务上的表现优于传统固定记忆方法。此研究为Agent领域的记忆系统设计提供了新的思路。意义：为Memory与Agent的结合应用开辟了新的研究路径。

OmniEcho: Spatial Audio Understanding for Embodied Agents

👍 22

09/19 20:00

Humans can effortlessly localize the direction of a sound source and integrate it with visual cues for reasoning, yet this remains challenging for embodied agents. In particular, it is still unclear how to effectively evaluate and model spatial audio understanding in embodied settings. To address th

中文介绍 OmniEcho研究了空间音频理解在具身智能体中的应用，提出了一种新的评估和建模方法。该方法能有效解析音源方向，并将其与视觉线索结合。针对现有模型在处理空间音频时的不足，提出了一种创新的算法框架，研究结果表明，智能体在复杂环境中对空间音频的理解和定位能力有了显著提升，推动音频智能研究的进展。意义：为Sound和Reasoning领域提供实用的解决方案。

Agent-Editing World Model: Rethinking World Modeling for LLM Agents

👍 19

09/22 20:00

Recent advances in large language models (LLMs) have enabled agents to tackle long-horizon tasks across diverse environments. To further improve agent performance, existing language world models typically predict environment observations, yet reconstructing high-entropy, execution-dependent tool res

中文介绍 针对现有的语言世界模型在高熵工具资源重建方面的不足，提出Agent-Editing World Model方法。该方法不仅可以预测环境观察，还能重构执行依赖的工具资源，从而显著提升智能体在复杂环境中的表现。实验结果表明，该模型在多样化长期任务中取得了较好的性能，推动了LLM在动态和不确定环境中的应用研究。意义：为World Model与Agent能力的提升提供关键支持。

Rufus-Air: An Open LLM Post-Training Recipe

👍 17

09/23 20:00

Rufus-Air is an open and reproducible post-training recipe on GLM-4.5-Air-Base (106B-A12B), organized as a serial pipeline of eight stages: SFT, Reasoning RL, Coding RL, Instruction-Following RL, General Agent, Coding Agent, Search Agent, and RLHF. We document the data, reward design, infrastructure

中文介绍 Rufus-Air提供了一种开放可重复的后训练配方，包含八个阶段，旨在优化大型语言模型（GLM-4.5-Air-Base）的后训练过程。通过详细记录数据、奖励设计及基础设施配置，实验展示了不同阶段对模型性能提升的积极影响。这一系统化的方法为后续研究提供了强有力的参考，推动了LLM领域的实用化进程。意义：提升了LLM在实际应用中的适应性和效率。

PACT: From Credit Assignment to Critic Alignment

👍 16

09/21 20:00

Reinforcement learning has become a central component of large language model (LLM) post-training, yet token-level credit lacks a generally accepted mathematical definition, leaving its relationship to commonly used training signals unclear. We formulate three regularity conditions, namely Completen

中文介绍 PACT研究重塑了强化学习中信用分配与批评者对齐的关系，提出了三种规律条件以建立清晰的数学定义。研究表明，传统的token级信用分配方式在实际应用中存在局限，通过此理论框架，能够更好地理解训练信号之间的关系。实验结果为强化学习中的优化策略提供了新的视角，促进了Agent智能化的进一步发展。意义：为Reinforcement Learning的理论研究提供了新方向。

Capable yet Parsimonious: Extracting and Characterizing Hidden Chain-of-Thought in Frontier Models

👍 15

09/21 20:00

The rapid capability gains of frontier language models are widely attributed to improved reasoning abilities, yet this cannot be verified as raw CoT traces in closed-source systems are hidden. By registering a simple custom tool through a standard API feature, we induce frontier models to externaliz

中文介绍 通过注册与API标准特性的自定义工具，提出了一种方法使前沿模型能够外化其思维过程。该研究揭示了这种思维链条的隐藏特征，并在多个任务中展示出更高的推理能力。实验结果表明，该方法可以有效提取和表征模型的潜在推理过程，为模型的透明性提供了利器，推动解释性AI的研究。意义：为Reasoning和Interpretability领域提供了重要信号。

Verifiable Hidden Dynamics Play: Generating Agentic RL Environments from Solved Mechanisms

👍 12

09/22 20:00

Language-model agents increasingly face long-horizon tasks with evolving state, interdependent decisions, and delayed outcomes. Scaling their training requires diverse agentic environments, dependable outcome signals, and low extension cost. Existing generation pipelines commonly construct an enviro

中文介绍 生成智能RL环境以应对模型在动态任务中的挑战，提出了一种新的生成管道，该管道能够构建多样化的任务场景，减少训练过程中的扩展成本。通过综合利用已解决机制，研究表明生成的环境能够有效支持智能体的长期任务训练，增强适应性与效果，推动RL领域的深入研究。意义：为Agent与World Model的结合提供了实用解决方案。

Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents

👍 11

09/23 20:00

The rapid progression of large language models is extending AI from passive content generation into the active workflows of engineering and scientific discovery. This shift raises a compelling question: can AI be both the object of development and an active participant in building next-generation AI

中文介绍 Qwen-Planner-Agent提供了一种封闭环的AI框架，旨在使智能体能够主动参与工程和科学发现的工作流中。通过解析当前LLM技术的大发展，该框架促进了AI从被动内容生成向主动参与的转变，展现了AI自主发展的潜力。研究表明，此方法在现实世界应用中具有广泛前景，为未来发展打下基础。意义：为AI与自动化协作的研究方向提供新思路。

PackLab: A Comprehensive Framework for Developing, Training, and Evaluating MLLMs in Robotic Bin Packing

👍 11

09/19 20:00

Robotic bin packing requires long-horizon sequential decision-making, as each object placement affects the available space for subsequent packing. Existing methods primarily rely on hand-crafted geometric heuristics that optimize predefined objectives or reinforcement learning policies learned throu

中文介绍 PackLab提供了一个全面的框架，用于开发、训练和评估机器人上的多层次模型（MLLM）在二次装箱任务中的表现。该框架克服了现有方法依赖手工几何启发式的局限，采用了强化学习与优化模型结合的策略，显著提高了多项决策性能。实验结果显示，模型在不同环境与约束条件下表现出色，为机器人任务规划研究提供了新的思路。意义：为Task和Motion Planning领域带来新的技术突破。

MemBodied: Recurrent Associative Memory for Vision-Language-Action Models

👍 11

09/22 20:00

Vision-Language-Action models provide a strong foundation for general-purpose robot control, yet a vast majority of policies do not preserve and leverage episode-level information beyond the current observation. This limitation is consequential in history-dependent manipulation tasks that depend on

中文介绍 MemBodied提出了一种递归关联记忆机制，旨在增强视觉语言行动模型在历史依赖性操作中的信息保持能力。通过改进记忆管理，该模型在多轮交互中显著提升了操作精确度，特别是在依赖于历史信息的场景中。实验表明，该方法在机器人控制与决策制定上展现出色，推动了多模态智能体的研究。意义：对Vision-Language-Action领域具有重要促进作用。

Learning to Discover Interesting Mathematics

👍 10

09/22 20:00

Recently, Large Language Models (LLMs) have been increasingly able to solve advanced mathematical problems, including many that have been open for decades. This opens the door to expansion of mathematical knowledge at unprecedented scale. Yet, while LLMs may be able to conjecture and prove more and

中文介绍 研究探讨了大型语言模型在数学问题求解中的潜力，尤其是解决长期未解问题的能力。透过新的推理机制，模型能够进行数学猜想与证明，为数学知识的扩展提供了前所未有的机会和规模。实验结果显示，模型在特定数学任务上具有较高的表现，加速了数学研究的进展与创新。意义：为AI在科学研究领域的应用提供了重要支持。

Coding Agents for Generalized Task and Motion Planning Problems

👍 10

09/23 20:00

Task and motion planning (TAMP) problems remain difficult even with full observability and object-centric states because discrete decisions are tightly coupled to geometric, kinematic, and dynamic constraints. Generalized TAMP addresses this difficulty by exploiting regularities across problem insta

中文介绍 针对任务与运动规划问题，提出了一种新的编码智能体架构，旨在更好地解决复杂决策的挑战，该模型通过利用问题实例之间的规律性，提高了规划效率。实验结果表明，这种方法在全局状态和离散决策相结合的情况下，显著提升了操作成功率，推动了规划及其应用的研究。意义：为Task和Motion Planning的研究提供了新的思路与方法。

Hunyuan-A13B Technical Report

👍 10

09/22 20:00

We present Hunyuan-A13B, an open-source large language model based on a Mixture-of-Experts architecture. It contains 80 billion total parameters but activates only 13 billion during inference, balancing model capability, computational efficiency, and deployment cost. The model is pretrained on a rig

中文介绍 Hunyuan-A13B是一个基于Mixture-of-Experts架构的开源大型语言模型，通过在推理时激活部分参数实现计算效率与能力的平衡。模型包含80亿参数，其特定激活策略提升了应用中的响应速度与资源利用率。实验结果表明，该模型在多种任务上的表现优异，为开发更高效的语言模型提供了实验依据。意义：为大规模模型的开发与应用提供了新方向。

IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis

👍 9

09/23 20:00

Deep search requires LLM agents to decompose complex queries, search for evidence, and synthesize grounded answers, yet existing ReAct-style agents suffer from two limitations: role coupling, where one policy must handle planning, evidence use, and synthesis; and context accumulation, where growing

中文介绍 IterSynth通过去耦角色的迭代合成方法，对深度搜索智能体进行重新审视。该方法解决了现有ReAct式智能体在计划、证据使用与综合过程中角色耦合的问题，通过优化上下文管理提高了搜索答案的准确性与效率。研究结果显示，通过限界条件与角色独立，能显著改善复杂查询的处理效果。意义：为Deep Search和AI智能体的改进提供了新思路。

The New Copilot

@jacobandreou · 7.8K 粉丝 · 521.0K 阅 · 500 赞 · 77 转

09/25 09:02

The bottleneck is no longer intelligence. Over the last six months, intelligence has accelerated dramatically. But smarter models are not enough for everyone to benefit from novel intelligence. New

中文介绍 随着智能模型的快速发展，过去六个月的人工智能技术趋于成熟。但仅依靠更聪明的模型无法让每个人受益，貌似新一轮的智能变革需要新的配套措施和理念。

We're open sourcing the company brain, Here's how we designed the multi-player harness

@DhravyaShah · 63.2K 粉丝 · 112.4K 阅 · 608 赞 · 45 转

09/25 22:08

It's been just 2 weeks since we discontinued the company brain harness. We had to offboard tons of people to other products, but many of our customers, and others suggested us to open source our work

中文介绍 在停用公司大脑助手后，博主讨论了开源设计多玩家架构的决定，旨在让更多人受益于他们的研究成果，回应了用户的强烈需求。

Why AI is booming, but productivity isn't

@chamath · 2.4M 粉丝 · 97.8K 阅 · 551 赞 · 44 转

09/25 14:06

In June, I wrote that vibe coding was dead and that ROI-driven analysis of AI was about to go from a nice-to-have to a necessity. Here's what AI ROI is and why it's tricky to see in GDP numbers so

中文介绍 博主指出，尽管人工智能领域正在蓬勃发展，但生产力却未必同步提升。他探讨了人工智能投资回报率的重要性和复杂性，认为未来是依靠数据驱动的投资决策。

Prescriptions for Prosperity in the Digital Economy

@saylor · 5.2M 粉丝 · 96.2K 阅 · 661 赞 · 98 转

09/26 09:11

Artificial intelligence will make it possible for individuals and companies to produce far more than they can today. That makes the freedom to create, finance, own, and exchange things more important.

中文介绍 人工智能将显著提高个人和企业的生产力，创造、融资、拥有和交换商品的自由将变得愈发重要。这一趋势反映了数字经济时代的潜力与挑战。

Jev is a state-of-the-art search reranker

@ErikKaum · 2.0K 粉丝 · 75.0K 阅 · 501 赞 · 44 转

09/24 13:06

We evaluated Jev as a reranker for search results in turbopuffer. The results may surprise you. Jev isn’t just a good reranker, it performed better than existing state-of-the-art rerankers in our

中文介绍 博主评估了 Jev 在搜索结果重排序中的表现，结果显示其效果优于当前顶尖重排序算法，揭示了新技术在搜索领域可能带来的突破。

Using Claude Code: Spending your effort

@trq212 · 354.7K 粉丝 · 62.7K 阅 · 965 赞 · 52 转

09/25 16:04

One of the best parts of our newest Claude models is how they respond to effort without breaking the prompt cache in Claude Code, but I’ve received a lot of questions on this from users. What is

中文介绍 利用 Claude Code 模型，通过合理设计提示优化用户的使用效率。博主回应了用户对新模型如何避免中断提示缓存的问题，阐述了其独特的响应机制。

We built a 20x faster Grok bot

@cerebras · 76.6K 粉丝 · 37.6K 阅 · 554 赞 · 28 转

09/25 11:28

Written by @milksandmatcha and @0xSero A personal assistant should save you time and effort. Over the past few weeks, we’ve been obsessively testing AI personal assistants on everyday tasks, from

中文介绍 团队测试出一种速度提升20倍的 Grok 个人助手，旨在节省时间和精力，通过 AI 完成日常任务，代表了个人助手领域的重要进展。

OpenRouter: from Seed to Stripe — with OpenRouter’s Alex Atallah & AMP’s Anjney Midha

Latent Space · 09/25 19:14

In 2023 most people doubted that there could be more than 1 or 2 frontier model labs. Now there are dozens.... and Stripe just bought the best known one for $7B.

中文介绍 2023年，许多人质疑前沿模型实验室的数量，但目前已有数十个实验室，而Stripe以70亿美元收购了知名的一个。

Proaction boosts sales 60% and saves 75+ hours with Codex

OpenAI News · 09/25 15:00

With Codex, GPT-Live-1, and GPT-6 Astra, Proaction builds, operates, and sells modern fleet management faster.

中文介绍 Proaction借助Codex实现销售增长60%，并节省超过75小时的工作时间，优化现代车队管理的构建、运营和销售。

The Pentagon wants $30 million to build an AI-powered lie detector

MIT Tech Review AI · 09/25 05:16

The US government wants to spend $30.3 million over the next five years on an improved form of lie detector, according to a Department of Defense budget request. The program, called Polygraph+ or Polygraph Next, will focus on scoring algorithms that use artificial intelligence and machine learning a

中文介绍 美国国防部提出预算申请，希望在未来五年内投入3030万美元开发一种改进版的测谎仪，名为Polygraph+，重点在于使用人工智能的评分算法。

[AINews] The Future of Latent Space

Latent Space · 09/25 01:37

A quiet day lets us discuss the work behind the scenes - now open for business!

中文介绍 今天是个安静的日子，我们讨论了幕后工作，现在已开放业务。

Runway’s WorldPrompt and the Engineering of Real-Time Worlds

Latent Space · 09/24 21:30

GWM Worlds 2 uses persistent context and timed actions to steer a world model generating video and audio in real time.

中文介绍 GWM Worlds 2利用持久上下文和定时操作来推动世界模型，以实时生成视频和音频。

ChatGPT Pro Max 🤖, Muse realtime avatar 🎭, DeepSeek $1B ARR 💰

TLDR AI · 09/24 20:00

中文介绍 ChatGPT Pro Max推出，Muse实时化身在使用中，DeepSeek的年收入达到10亿美元。

Introducing Gemini 3.8 Live with Live Avatar

DeepMind Blog · 09/24 12:20

中文介绍 DeepMind推出Gemini 3.8 Live及其实时化身功能。

Foundries vs Navigators: Lowering the Cost of Science

Latent Space · 09/24 11:03

Guest Post: In science, thinking has gotten cheap but doing has not. This asymmetry is reshaping how research companies operate, largely inconspicuously.

中文介绍 在科学领域，思考成本降低，但实际执行成本仍高，这种不对称正在重塑研究公司的运作方式。

Accelerating vision-language models with LFM2.5-VL-DSpark

Hugging Face Blog · 09/24 10:08

中文介绍 通过LFM2.5-VL-DSpark加速视觉-语言模型的发展。

[AINews] Meta Connect 2026: Muse glasses, voice, video, and Charm

Latent Space · 09/24 04:12

Team Zuck is absolutely on fire.

中文介绍 Meta团队表现积极，推出了Muse眼镜、语音、视频等新技术。

Gemini TTS 🗣️, Claude’s novel enzyme 🧬, Google private memory 🔒

TLDR AI · 09/23 20:00

中文介绍 Gemini TTS在语音合成领域有所进展，Claude推出新型酶，Google增强了私人记忆功能。

How to Use NVIDIA Warp and MjWarp to Accelerate Robotics Simulation and Learning Workflows

Hugging Face Blog · 09/23 14:41

中文介绍 介绍如何使用NVIDIA Warp和MjWarp加速机器人模拟和学习流程。

Advancing Private AI Compute with secure, server-side memory

DeepMind Blog · 09/23 12:00

Introducing private, server-side memory to Private AI Compute for personal AI.

中文介绍 DeepMind推出安全的服务器端内存，以支持个人人工智能计算的隐私。

Two years of OpenAI Academy

OpenAI News · 09/23 12:00

Marking two years of OpenAI Academy and bringing AI skills to even more communities.

中文介绍 OpenAI Academy庆祝成立两周年，致力于将人工智能技能带给更多社区。

Gemini 3.8 text-to-speech says hello

DeepMind Blog · 09/23 11:25

中文介绍 Gemini 3.8语音合成技术正式推出。