2026-09-26

Training Object Permanence in World Models

👍 186

09/23 08:00

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 针对视频生成模型是否具备物体永久性与固体性等人类认知先验，论文提出在其中训练物体永久性能力，考察并增强世界模型对遮挡与物体持续存在的推理，推动更接近人类的物理智能。

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

👍 82

09/22 08:00

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 针对多方对话的长期记忆，提出以说话人为中心的双轨记忆框架 SpeakerMem-R1，不仅检索相关内容，还区分谁说了什么、每句话涉及谁、个体间的相互认知、群体共享信息及状态随时间的变化。

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

👍 55

09/24 08:00

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 发现大语言模型存在基础线性性质：当来自不同文本流的输入被线性组合时，模型输出为各自 next-token 分布的叠加，作者称之为 linear superposition，为理解 LLM 内部并行信息表征提供证据。

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

👍 45

09/19 08:00

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 面向视觉语言模型在物理世界中的空间推理，提出 Spatial-Interactor，让模型通过与可观测环境的交互感知物体运动与视角变化引起的局部状态转移，并在长轨迹上整合以维护更新的世界状态。

HappyWorld-Bench

👍 40

09/21 08:00

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 提出 HappyWorld-Bench 综合基准，用于评测世界模型所生成世界的质量，以及在被智能体探索、交互和修改时的一致性与响应能力，弥补现有评测偏重生成质量而忽视可控性的不足。

The Past Frames the Future: Memory for Autoregressive Video Generation

👍 38

09/23 08:00

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 针对自回归视频生成中随序列延长而累积的误差与一致性退化瓶颈，论文引入记忆机制，用过去帧信息约束后续帧生成，以支撑长时程生成、交互式世界建模与持续演化的视觉环境。

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

👍 34

09/23 08:00

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 现有 agent 记忆系统多在写入时固化，把任务轨迹蒸馏为反思、工作流或技能等固定产物再检索。论文提出 Just-in-Time Memory，让 LLM agent 在任务执行时按需整理任务自适应记忆。

OmniEcho: Spatial Audio Understanding for Embodied Agents

👍 20

09/20 08:00

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 针对具身智能体难以定位声源方向并与视觉线索联合推理的问题，论文提出 OmniEcho，构建评测基准与建模方法，系统衡量并提升具身场景下的空间音频理解能力。

PACT: From Credit Assignment to Critic Alignment

👍 16

09/22 08:00

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 指出 token 级 credit assignment 缺乏公认数学定义，提出完整性等三条正则条件以厘清其与常用训练信号的关系，并据此给出从信用分配到 critic 对齐的方法 PACT。

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

👍 14

09/22 08:00

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 针对闭源前沿模型隐藏原始 CoT、推理能力无法验证的问题，作者通过标准 API 注册自定义工具，诱导模型外化隐藏的思维链，从而抽取并刻画其推理过程与能力边界。

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

👍 13

09/23 08:00

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 现有语言世界模型多预测环境观测，但高熵且依赖执行结果的工具返回难以重建。论文提出 Agent-Editing World Model，把世界建模重构为对状态的编辑式更新，以更好支撑 LLM agent 的长时程任务。

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

👍 10

09/20 08:00

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 面向机器人装箱这一长时程序列决策任务，提出 PackLab 框架，用于开发、训练与评测多模态大模型，以替代依赖人工几何启发式或强化学习策略的传统做法。

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

👍 10

09/23 08:00

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 针对视觉-语言-动作模型普遍只依赖当前观测、无法利用片段级历史的问题，提出 MemBodied，用循环联想记忆为 VLA 策略保留并调用历史信息，改善依赖历史的操作任务表现。

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

👍 9

09/24 08:00

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 提出 Qwen-Planner-Agent 闭环 AI-for-AI 框架，让大模型既是开发对象、也是构建下一代移动规划智能体的参与者，在真实移动场景中实现规划能力的自动迭代与提升。

Hunyuan-A13B Technical Report

👍 9

09/23 08:00

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 开源大模型 Hunyuan-A13B 采用 MoE 架构，总参数 80B，推理时仅激活 13B，在模型能力、计算效率与部署成本之间取得平衡，并基于一套严谨的预训练数据与流程完成训练。

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

👍 8

09/24 08:00

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 针对 deep search agent 中角色耦合与上下文累积两大问题，提出 IterSynth，通过角色解耦的迭代合成，将规划、证据使用与答案综合分离，提升复杂查询的证据检索与答案可靠性。

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

👍 8

09/23 08:00

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 为扩展语言模型 agent 的长时程训练，提出 Verifiable Hidden Dynamics Play，从已解机制的谜题中生成可验证的 agentic RL 环境，提供可靠的结果信号并降低环境扩展成本。

WhatWorkedBench: Benchmarking Experimental Understanding in AI Agents

👍 8

09/23 08:00

AI research agents need reliable knowledge of how their experiments change outcomes. We introduce WhatWorkedBench to measure experimental understanding, the accuracy of predictions about component changes after budgeted experimentation. Agents inspect code, select measurements, and submit a response

中文介绍 提出 WhatWorkedBench，衡量 AI 研究 agent 的实验理解能力：在有限实验预算下检查代码、选择度量后，预测组件改动将如何影响结果，据此评估其科研实验判断力。

Rufus-Air: An Open LLM Post-Training Recipe

👍 7

09/24 08:00

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 是基于 GLM-4.5-Air-Base（106B-A12B）的开放可复现后训练配方，串联 SFT、推理 RL、代码 RL、指令遵循 RL、通用与代码及搜索 Agent、RLHF 共八个阶段，并公开数据与奖励设计细节。

RGBD20K: A Large-Scale Benchmark for RGB-D Semantic Segmentation

👍 6

09/24 08:00

In this paper, we propose RGBD20K, a novel dataset for facilitating the development of more robust and general RGB-D semantic segmentation by encompassing abundant categories and high-quality annotations. RGBD20K possesses several attractive properties: (1) Expanded Semantic Space. In particular, it

中文介绍 提出 RGB-D 语义分割大规模数据集 RGBD20K，涵盖丰富类别与高质量标注，显著扩展语义空间，用于推动更鲁棒、更通用的 RGB-D 语义分割模型发展。

How To Code Almost Forever for $20/Month With Codex + GPT-6 Luna

@sairahul1 · 147.5K 粉丝 · 785.8K 阅 · 506 赞 · 61 转

09/23 17:41

Yesterday OpenAI released GPT-6 Luna. And it might completely change the economics of AI coding. Here is what most people are missing. $20/month ChatGPT Plus plan with Codex now gives you access to

中文介绍 OpenAI 发布 GPT-6 Luna 后，作者算了一笔账：ChatGPT Plus 每月 20 美元配合 Codex，即可近乎无限地写代码，认为这改变了 AI 编程的成本结构。文中指出多数人忽略的订阅档位与用量细节，属于对现有性价比的重新解读。

How to Build Agentic Harness using Jev (Builder's Guide)

@Av1dlive · 26.6K 粉丝 · 339.1K 阅 · 515 赞 · 61 转

09/24 00:49

i’m going to walk you through how i built an agentic harness with jev, where i put the decision layer....step by step, with the code and diagrams behind each choice. TLDR; if you don't wanna read all

中文介绍 一份用 Jev 搭建 agentic harness 的实战指南，重点讲决策层放在哪里，为什么这样分层。作者逐步拆解每个设计选择，并附上对应代码与架构图，适合想把 agent 从 demo 推到可用工程结构的开发者。

How to automate SEO with Opus 5.5 (Full Course)

@EXM7777 · 146.5K 粉丝 · 316.1K 阅 · 552 赞 · 33 转

09/23 21:54

I'm going to walk you through the SEO & AEO setup i run with AI agents, from the lazy autopilot version to the full Claude Code build on Opus 5.5, step by step... because search changed shape a while

中文介绍 一套用 AI agent 做 SEO 与 AEO 的完整配置教程，从最省事的全自动版本，一路讲到基于 Opus 5.5 的 Claude Code 自建方案。作者按步骤拆解，前提是搜索的分发逻辑已经变了，旧打法需要重做。

The $4 Trillion Revolution: How AI Agents Will Rewire Finance and Spark an Economic Big Bang

@joechalom · 18.9K 粉丝 · 105.3K 阅 · 505 赞 · 113 转

09/23 20:00

The convergence of new onchain financial primitives and unlimited agent attention promises unprecedented value creation for consumers. The battle over control has begun. Over $1 trillion of annual

中文介绍 观点长文：链上金融原语与 agent 无限注意力结合，可能带来四万亿美元级价值重估。作者判断消费者端会先受益，而围绕控制权的争夺已经开始，涉及每年超一万亿美元的存量市场，属于宏观叙事而非实操。

Why AI is booming, but productivity isn't

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

09/26 02:06

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 作者延续六月「vibe coding 已死」的判断，讨论为何 AI 投资火热却看不到生产率提升。核心是拆解 AI ROI 该怎么算，以及为什么它很难体现在 GDP 数字里，是偏经济学视角的冷静分析。

How to train your own Jev for $17

@nutlope · 100.4K 粉丝 · 89.8K 阅 · 518 赞 · 35 转

09/24 06:02

TLDR: We just launched our own Jev-like classifier, together/Tev1-4B-experimental, on top of Qwen3.5 4B. In this blog post we’ll show you how to fine-tune your own version! Jev has quickly become one

中文介绍 团队在 Qwen3.5 4B 上微调出 Jev 风格的分类器 together/Tev1-4B-experimental，并公开完整博客，教你自己花 17 美元训一个。适合想低成本复现小模型微调流程、又不想从零摸索的人。

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

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

09/25 01:06

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 作者在 turbopuffer 上把 Jev 当作搜索结果重排器做评测，结论是它不仅好用，还超过了现有的 SOTA 重排模型。对于正在选型搜索 reranker 的团队，这是一份可直接参考的对比结果。

Using Claude Code: Spending your effort

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

09/26 04:04

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 Claude 新模型的一大改进是响应 effort 参数时不会破坏 Claude Code 的 prompt cache，作者针对用户集中提问做了解释。内容偏机制说明，讲清 effort 该怎么花、为什么会省下重复计费。

Building a Custom Harness with Pi and Jev

@omarsar0 · 321.0K 粉丝 · 45.0K 阅 · 555 赞 · 46 转

09/23 22:09

An AI agent is a language model that works in a loop. It reads the task, uses a tool such as "read this file" or "delete that file", looks at the result, and keeps going until the job is done. Each

中文介绍 从 agent 的基本形态讲起——模型读任务、调工具、看结果、循环直到完成——再进入用 Pi 和 Jev 搭自定义 harness 的实践。适合想理解 agent 循环本质并动手改造自己脚手架的读者。

Gemini 3.8 text-to-speech says hello

@GoogleAIStudio · 205.4K 粉丝 · 41.9K 阅 · 519 赞 · 50 转

09/24 00:20

Today, we’re introducing two new text-to-speech models to the Gemini family, transforming voice generation from static presets into a dynamic creative studio. These models enable creators, developers,

中文介绍 Google 在 Gemini 家族下新增两个文本转语音模型，把语音生成从固定预设推进到可动态调教的创作工作台。面向创作者与开发者，具体能力与接入方式在推文中列出，属于官方产品发布。

We built a 20x faster Grok bot

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

09/25 23:28

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 Cerebras 团队用自家硬件搭了一个速度提升 20 倍的 Grok 机器人，作为个人助理跑日常任务。文章记录了连续数周压测各类 AI 助理的实际表现，角度是推理速度对助理体验的真实影响。

Ghast AI Season 3 Starts September 25

@Ghast_AI · 10.0K 粉丝 · 6.3K 阅 · 542 赞 · 31 转

09/24 22:29

We just introduced the Capability Economy, a market where private AI work can be bought and verified without exposing the method behind it. Every Capability in that market starts as a workflow someone

中文介绍 Ghast AI 第三季定档 9 月 25 日，同时抛出「能力经济」概念：一个可以买卖并验证私有 AI 工作成果、却不暴露方法本身的市场。每个能力都源自某人的工作流，属于新机制预告。

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

Latent Space · 09/26 07:14

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 播客对话OpenRouter的Alex Atallah与AMP的Anjney Midha，回顾该公司从种子轮走到被Stripe收购的历程。2023年多数人怀疑前沿模型实验室只会有一两家，如今已有数十家，Stripe以70亿美元收购其中知名度最高的一家。

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

OpenAI News · 09/26 03:00

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

中文介绍 OpenAI介绍企业Proaction借助Codex、GPT-Live-1与GPT-6 Astra，更快地构建、运营和销售现代车队管理业务，销售额因此提升60%，并节省75小时以上工时。

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

MIT Tech Review AI · 09/25 17:16

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 据美国国防部预算申请，五角大楼计划未来五年投入3030万美元研发改进型测谎技术，项目名为Polygraph+（又称Polygraph Next），重点开发基于人工智能的评分算法。

[AINews] The Future of Latent Space

Latent Space · 09/25 13:37

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

中文介绍 Latent Space发布AINews，称当日新闻较为平淡，借此介绍团队幕后工作进展，并宣布相关业务现已对外开放。

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

Latent Space · 09/25 09:30

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 Runway介绍WorldPrompt与实时世界生成技术。其GWM Worlds 2通过持久上下文和定时动作控制世界模型，实时生成视频与音频。

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

TLDR AI · 09/25 08:00

中文介绍 TLDR AI当日简报要点：ChatGPT Pro Max、Muse实时数字人，以及DeepSeek年度经常性收入达到10亿美元。

Introducing Gemini 3.8 Live with Live Avatar

DeepMind Blog · 09/25 00:20

中文介绍 DeepMind发布Gemini 3.8 Live，并新增Live Avatar实时数字人功能。

Foundries vs Navigators: Lowering the Cost of Science

Latent Space · 09/24 23:03

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 Latent Space刊发客座文章，指出科学研究中思考的成本已变得低廉而执行成本依旧高昂，这种不对称正在悄然重塑研究型公司的运作方式。

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

Hugging Face Blog · 09/24 22:08

中文介绍 Liquid AI在Hugging Face博客发布LFM2.5-VL-DSpark，用于加速视觉语言模型的运行与推理。

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

Latent Space · 09/24 16:12

Team Zuck is absolutely on fire.

中文介绍 Latent Space的AINews汇总Meta Connect 2026发布内容，包括Muse智能眼镜，以及语音、视频和Charm相关功能。

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

TLDR AI · 09/24 08:00

中文介绍 TLDR AI当日简报要点：Gemini文本转语音、Claude参与发现的新型酶，以及谷歌推出的私密记忆功能。

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

Hugging Face Blog · 09/24 02:41

中文介绍 Hugging Face博客介绍如何使用英伟达Warp与MjWarp加速机器人仿真与学习工作流。

Advancing Private AI Compute with secure, server-side memory

DeepMind Blog · 09/24 00:00

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

中文介绍 DeepMind宣布为Private AI Compute引入私有服务器端内存，面向个人AI场景，以推进隐私保护的AI计算。

Two years of OpenAI Academy

OpenAI News · 09/24 00:00

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

中文介绍 OpenAI Academy成立两周年，该计划旨在把AI技能带给更多社区。

Gemini 3.8 text-to-speech says hello

DeepMind Blog · 09/23 23:25

中文介绍 DeepMind发布Gemini 3.8文本转语音功能。