2026-09-05

Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments

👍 229

09/02 20:00

As terminal-based code agents become prevalent, agent trajectories have accumulated at scale, while realistic, executable environments remain scarce. However, environments are what agent post-training actually requires: each can be re-queried into many verifiable tasks and provides execution feedbac

中文介绍 该论文提出了一种将智能体轨迹转化为可扩展终端环境的方法，以应对当前可执行环境稀缺的问题。通过合成逼真的环境，论文使得每个环境都可重新查询为多个可验证任务，从而提供执行反馈，显著提升智能体的后训练性能。意义：为智能体的学习与适应性提供更为有效的环境支持。

Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning

👍 159

09/02 20:00

Large language models achieve superior performance on tasks that require extended reasoning, but long chains of thought make the KV cache a severe memory bottleneck. Existing KV cache compression methods share one paradigm: score each cached token by some estimate of how much it will matter later, a

中文介绍 论文探讨了在需要延续推理的任务中，如何通过随机注意机制改进KV缓存驱逐策略，以解决大型语言模型中的内存瓶颈问题。提出的方法打破了传统缓存压缩策略的范式，着重于动态评估被缓存令牌的重要性，从而实现更高效的推理，具有潜在的广泛应用价值。意义：推动了大型模型推理效率的提升，可能影响推理与生成类应用。

Knowing When Not to Reuse: Conditional Experience Transfer in Autonomous LLM Post-Training

👍 149

08/26 20:00

Large language models offer broad capabilities, but adapting them to evolving domains, tools, and requirements often entails repeated post-training. Autonomous systems automate parts of this process by proposing updates, training candidates, and using evaluation feedback to select subsequent proposa

中文介绍 本研究提出了一种自主更新方法，通过条件经验转移来优化大型语言模型在不断变化的领域中的适应能力。该方法通过自动提议更新和使用评估反馈来选择后续提案，旨在减少手动重训练的需求，提高模型的灵活性和实用性，具有重要的实用意义。意义：推动了自主系统与动态环境适应性研究的发展。

Rethinking On-Policy Distillation of Large Language Models II: One Training Example

👍 73

09/02 20:00

On-policy distillation (OPD) combines student-generated rollouts with dense token-level supervision from a teacher. Existing work has mainly studied its algorithmic behavior, leaving the role of training data unclear. We examine this role at the data-minimal limit by training on a single query. One-

中文介绍 研究探讨了在仅使用单个查询的情况下，On-policy distillation（OPD）对大型语言模型进行的直接提取方式。通过减少对训练数据的依赖，论文阐明了数据在模型训练中的重要性，为即便在数据稀缺情况下的高效模型训练提供了新思路。意义：为模型压缩和高效训练方法提供了新的范式，增强了模型在资源有限环境下的适应能力。

Why Gated DeltaNet Survives 4-Bit Quantization: NVFP4 W4A4 for the Recurrent Half of a Hybrid 27B LLM

👍 73

09/02 20:00

Hybrid LLMs pair softmax attention with linear-attention layers such as Gated DeltaNet (GDN), whose recurrent state summarizes the context in fixed size. Early community 4-bit quantizations of Qwen3.8-27B (48 GDN layers, 16 attention layers) left the GDN block in 8- or 16-bit precision -- especially

中文介绍 论文分析了Gated DeltaNet在4位量化下的性能表现，指出该模型在混合大型语言模型中的状态保持固定大小的重要性。通过实验证实该模型在低精度量化下依然保持较高的性能，提出了对未来模型设计的启示，强调了量化和效率之间的平衡。意义：推动了大型模型在硬件资源受限场景中的性能优化研究。

Scal3R: Learning Efficient Multi-Relative Pose Query for Scalable Online 3D Reconstruction

👍 45

09/02 20:00

Online 3D reconstruction models perform poorly on long videos. This happens because regressing poses relative to a fixed first-frame anchor forces extrapolation far beyond the training distribution. Small drifts accumulate and amplify into significant geometric collapse. However, we observe that per

中文介绍 本研究针对在线3D重建提出了Scal3R模型，以解决长视频中的重建效果不佳的问题。通过回归相对于第一个帧的动态姿态，避免了小的漂移累积导致的几何崩溃，提出了一种新颖的多相对姿态查询方法，从而提升模型在复杂场景中的稳定性和准确性。意义：为实时3D重建技术的发展提供了新的方向和可能性。

Editable Visual Design

👍 39

09/02 20:00

While diffusion base models such as GPT-Image-2 and Nano-Banana exhibit remarkable visual expressiveness, their end-to-end generation inherently yields flattened bitmaps with error-prone text, precluding layer-wise post-editing. Conversely, code-based visual generation via Coding Agents provides pre

中文介绍 论文提出了一种可编辑的视觉设计方法，克服了传统扩散模型生成的图像易出现错误文本的问题。通过基于代码的视觉生成，研究者实现了图层级别的后编辑能力，为视觉内容创作提供了更大的灵活性和有效性。意义：推动了可编辑生成内容的研究，为图像生成与编辑流程提供新思路。

Beyond Retrieval: Progressive Latent Memory Evolution for Streaming Video Understanding

👍 29

09/02 20:00

Streaming video understanding requires multimodal large language models (MLLMs) to process continuous visual inputs and respond to user queries under strict causality and bounded memory. Existing approaches typically compress historical observations into an external memory bank and retrieve query-re

中文介绍 该研究探索在流媒体视频理解中如何利用多模态大型语言模型（MLLM）进行连续视觉输入的处理。通过压缩历史观察数据和外部记忆库的结合，论文旨在提升模型的因果推理能力和内存效能，推动了多模态理解技术的发展。意义：为视频理解和检索系统的效率提升提供了新的思路和方法。

WHALE: A Simple Recipe for Joint Harness-Weight Optimization

👍 29

08/30 20:00

Agent performance depends jointly on the model parameters and the executable harness code that manages context and control flow. Optimizing either component in isolation can leave the system bottlenecked by its frozen counterpart: weight updates can change which harness is effective, while harness u

中文介绍 论文介绍了WHALE方法，用于联合优化智能体的模型参数和执行代码，以提升整体性能。研究表明，隔离优化将导致性能瓶颈，并提出了动态更新策略的有效性，从而确保智能体在广泛的环境中保持高效的决策能力。意义：推动智能体训练和优化流程的发展，提升智能体在复杂场景中的适应性。

Beyond Visual Similarity: Entity-Aligned Retrieval for Knowledge-Based Visual Question Answering

👍 26

08/18 20:00

Knowledge-Based Visual Question Answering (KB-VQA) relies on retrieving external information to answer queries involving long-tail entities. However, existing retrieval pipelines predominantly employ CLIP-style dual encoders, which prioritize surface-level visual similarity over entity-level semanti

中文介绍 该研究针对知识基础的视觉问答（KB-VQA）中，实体对齐的检索问题提出新的解决方案。通过引入基于实体的语义相似性，论文旨在提升在复杂查询中对于长尾实体的识别能力，改善现有检索的准确性和可靠性。意义：为视觉问答领域推动基于知识的检索技术进步提供了新思路。

RealSWE: A Compositional Evaluation of Coding Agents under Realistic User Requests

👍 24

08/30 20:00

Coding agents are now commonly evaluated on the SWE-bench family of benchmarks, whose tasks are built from curated GitHub issues: long, structured, and information-rich. Real user requests, however, are typically far shorter and less structured. To characterize this gap, we define a six-category inf

中文介绍 研究分析了真实用户请求与现有编码代理评估基准之间的差距，提出了六类信息请求来体现这一问题。通过探索短而结构化的请求特征，论文为编码代理的发展提供了新的评估视角，强调了真实世界应用中的有效性和适应性。意义：推动编码代理技术在实际应用中的发展与完善。

CORE: Improving Compositional Reasoning in MLLM Embedding via Reranker Distillation

👍 24

09/02 20:00

MLLM-based embedding models remain limited in compositional retrieval, often failing to distinguish scenes containing the same concepts but different attribute-object bindings. Yet the same backbone can resolve such distinctions when used as a cross-attentive reranker, motivating us to distill its c

中文介绍 该论文研究了如何通过Reranker Distillation改善多模态大型语言模型的组合推理能力，尤其是在相同概念但不同属性对象排列的场景中。研究结果表明，利用跨注意力机制的reranker能够有效提升模型在复杂检索任务中的性能，为未来的检索优化提供了指引。意义：推动了组合推理研究的发展，提升了检索准确率。

ZipTok3D: High-Fidelity 3D Tokenization with Compact Token Prefixes

👍 24

08/31 20:00

Compact token sequences are essential for efficient 3D generation. However, existing 3D tokenizers typically organize latent representations either over spatial regions or as fixed-size sets of global tokens, both suffering sharp reconstruction degradation when compressed to extremely low token budg

中文介绍 论文提出ZipTok3D模型，针对高保真3D生成中的紧凑令牌化问题进行研究。通过改进的令牌生成策略，该模型有效减少了压缩带来的重建质量下降问题，为实现高效3D生成提供了新方法。意义：推动3D生成技术的提升与应用普及。

DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training

👍 23

09/02 20:00

Reinforcement Learning from Verifiable Rewards works well when a task has a programmatic checker, but most long-horizon agent domains have none. We work in the outcome-blind setting, where ground-truth success signals are not available. Multi-criteria rubrics are a popular way to supply such a rewar

中文介绍 研究提出了DRACO方法，旨在通过动态标准提升长时间智能体训练中的信用分配效率。该方法适应于缺乏真实成功信号的环境，通过多标准评估提升了反馈策略的有效性，为长时间持续学习问题提供解决方案。意义：推动强化学习领域中的长时间任务训练系统的发展。

WorldReward: Reward Modeling for Camera-Conditioned World Models

👍 23

09/02 20:00

Camera-conditioned world models generate interactive videos in which commanded actions should induce the expected scene changes while appearance, geometry, and temporal dynamics remain coherent. Existing rewards assess these requirements separately: geometry-based rewards estimate trajectory executi

中文介绍 本研究提出了一种新的奖励建模方法WorldReward，专为相机条件下的世界模型设计。通过综合几何、外观和动态时间变化等因素，该方法致力于处理生成视频中的物理推理挑战，以提升模型在任务中的表现。意义：增强了世界模型领域的研究，为交互式视频生成的精确控制提供支持。

PACE: Towards Surfacing Hidden Conflicts in User Requests

👍 22

09/02 20:00

Personalized assistants should not only comply with user requests but also assess whether those requests are appropriate given the user's current circumstances. However, prior work has primarily focused on accurately executing requests, overlooking the need for assistants to account for context and

中文介绍 论文探讨了个性化助手在执行用户请求时，如何评估请求的适当性。研究提出了一种动态上下文判断机制，以确保助手不仅能准确执行请求，还能考虑用户的实际情况，为智能助手的发展提供了新的视角。意义：推动个性化服务与上下文感知技术的结合发展。

Using Grounded Theory for Agent Behavior Analysis at Scale

👍 18

08/30 20:00

Understanding agent behavior requires methods that scale to thousands of trajectories and surface new patterns in long, often unfamiliar tasks where pre-built classifiers fall short. We propose to bring grounded theory into agent trajectory analysis: a six-decade-old qualitative method from the soci

中文介绍 该论文提出将扎根理论应用于智能体行为分析，以应对传统分类器在分析大量轨迹时的局限性。通过量化和定性的方法，研究展示了在复杂任务中发现新模式的可能性，努力提升智能体行为理解的深度与广度。意义：为智能体行为分析的定量与定性结合提供新方法。

Environment Evolution for Terminal Agents

👍 17

09/02 20:00

Scaling interactive and verifiable environments is critical for training terminal agents. As frontier models become more capable, environments synthesized from scratch become less challenging and thus provide limited learning signals. Recent co-evolution methods iteratively synthesize environments n

中文介绍 论文提出了环境演化方法，以应对终端智能体训练中环境不足的问题。研究表明，通过协同进化合成更具挑战性的环境，能有效激励模型学习，从而提升智能体的适应能力，为强化学习环境的有效利用提供理论支持。意义：推动终端智能体训练的环境设计与优化研究。

Principia: Relational Physics Tests for Video Models

👍 16

09/02 20:00

Evaluating physical reasoning in video models is difficult because absolute motion measurements depend on frame rate, object scale, and camera calibration, all of which are often ambiguous or unavailable in generated video. We propose a different approach. When two objects in the same scene obey the

中文介绍 研究提出Principia，解决视频模型中物理推理精度评估的问题。通过引入新的相对运动测量方法，该方法有效处理视频生成中的帧率、物体尺度等不确定因素，提高模型的物理推理能力，为视频理解技术发展提供新思路。意义：推动视频智能理解领域的精确推理研究进展。

Select, Compress, Reinvest: A Controlled Study of Visual-Token Allocation in Long-Video MLLMs

👍 15

09/02 20:00

Long-video language models cannot look at every frame: an hour sampled once per second is 3,600 images, and a system keeps only a small fixed slice of that pool. Which frames survive that slice is usually treated as a preprocessing detail; we test whether it should be. Published selectors make the c

中文介绍 论文探讨了在长视频多模态语言模型中，如何有效选择与压缩视觉令牌。通过对选择过程进行系统研究，研究者分析了选择对模型性能的影响，提供了对视觉信息处理有效性的深刻见解，旨在提升长视频理解和分析的效果。意义：为长视频分析与处理的优化研究提供新的视角和方法。

Grok Imagine Odyssey Contest Winners

@imagine · 79.2K 粉丝 · 714.5K 阅 · 532 赞 · 86 转

09/04 16:49

Homer had a lyre. You had Grok Imagine. Thousands of you quoted our launch post and built original Odyssey scenes in Grok Imagine. Image, video, and the voices. Three to five minutes, with at least

中文介绍 分享了Grok Imagine比赛的获胜者和参与者的创意场景，展示了如何利用平台生成三到五分钟的视觉和音频内容。这反映了用户在使用Grok Imagine过程中展示创造力的趋势。

Five things I learned from using Astra

@dkundel · 23.9K 粉丝 · 225.2K 阅 · 520 赞 · 44 转

09/04 16:27

Back in February, getting Codex to build LEGO models took a custom skill, complicated instructions, and hours of iteration. And it was still not great. I came away thinking it wasn’t something models

中文介绍 讲述了在使用Codex构建LEGO模型过程中的挑战，包括需要复杂指令和多次迭代才能实现目标。强调了AI工具在创造性任务中的局限性，揭示了开发者的实际体验。

CarryGo, here we go!

@CarryGo_AI · 9 粉丝 · 192.8K 阅 · 1.0K 赞 · 2 转

09/04 06:34

Inside CarryGo lies a world model—helping organizations see, simulate, act and learn as a system. At the product level, CarryGo is a world model, built first for business. At the system level, it is

中文介绍 介绍了CarryGo作为一个世界模型的产品，旨在帮助组织进行系统性的学习与决策。此工具是针对企业需求设计的，反映了商业领域对系统思维的日益重视。

Prediction: AI will collapse

@wordgrammer · 26.1K 粉丝 · 120.8K 阅 · 602 赞 · 33 转

09/02 17:30

The dot com bubble got some things right. It noticed that the internet was going to be big. It noticed that the value generated by the internet would roughly scale in proportion with the amount of

中文介绍 讨论了AI产业的未来可能会经历崩溃，借鉴互联网泡沫的历史经验，提醒人们对AI生成价值的过度乐观。此观点引发对行业持续泡沫的担忧。

Applied AI Doesn't Work

@vasuman · 96.1K 粉丝 · 109.9K 阅 · 517 赞 · 32 转

09/04 18:17

The world has spent a fortune on AI. For most enterprises, nothing changed. I've spoken to over 300 CEOs, CIOs and CFOs at the largest companies on Earth. I learned most companies apply AI onto their

中文介绍 强调尽管全球在AI上投入巨额资金，企业采纳AI的效果却甚微。通过与300多位高管的交谈，揭示了企业在应用AI过程中的普遍障碍，反映出行业对AI落地效果的质疑。

AI Engineering Skills Map: Using coding agents

@AndrewYNg · 1.9M 粉丝 · 48.9K 阅 · 918 赞 · 136 转

09/04 11:02

A key AI engineering skill is using coding agents. Your skill at steering them both to write code and to carry out non-code tasks, such as analyzing data or managing system operations, allows you to

中文介绍 分享了AI工程的重要技能，即利用编码代理进行代码编写和数据分析。这种技能的提升能够显著提高工作效率，反映了对多样化AI应用能力的重视。

Rethinking skills and prompts for GPT-6 Astra

@pvncher · 35.4K 粉丝 · 44.6K 阅 · 508 赞 · 37 转

09/04 17:44

Coding agents have come a long way, and best practices are changing fast. What used to require a lot of handholding and scaffolding no longer does. If you’ve been using agents in your projects over

中文介绍 探讨了编码代理的最佳实践快速演变，指出当前的需求不再需要过多的指导，这标志着AI应用的成熟与简化，显示出行业对提高AI使用便利性的共识。

Grok Bot for Work

@joshkim · 8.3K 粉丝 · 42.4K 阅 · 525 赞 · 41 转

09/03 18:03

I can't do my job without Grok Bot. It happened really quickly. A few weeks ago, I wasn’t leveraging agents like the AI-pilled experts you see on X and LinkedIn. I was still a human in the loop:

中文介绍 分享了个人在工作中对Grok Bot的依赖，强调其在提高工作效率方面的重要性，反映出越来越多的专业人士正在积极拥抱AI代理工具。

Codex has changed my life as a solopreneur

@jonnym1ller · 20.7K 粉丝 · 36.0K 阅 · 510 赞 · 23 转

09/03 18:53

It's difficult to exaggerate how absurdly life-changing Codex has been for me as a solopreneur in recent months. And yet people have been saying 'but what do you actually use it for' (so I decided to

中文介绍 描述了Codex对作为独立企业家的极大影响，分享了具体应用场景以回应他人的质疑。这表明越来越多的创业者正借助AI工具实现业务转型与效率提升。

Introducing Gemini 3.8 Flash and 3.8 Flash Cyber

@GoogleAIStudio · 198.5K 粉丝 · 26.2K 阅 · 507 赞 · 59 转

09/02 12:48

Our newest Gemini models deliver next-generation intelligence for agentic workflows and cybersecurity. Building on the momentum of 3.7 Flash from three weeks ago and marking our third Flash release in

中文介绍 推出了Gemini 3.8 Flash和3.8 Flash Cyber，展示了新一代智能在代理工作流和网络安全中的应用。这一更新表明AI技术持续快速进步及其在安全领域的潜在影响。

OpenClaw Power, MacBook Simplicity: Five Days With Grok Bot

Latent Space · 09/05 11:01

SpaceXAI’s Grok Bot has the same level of programming power as OpenClaw, but it’s programmable at a different level of abstraction.

中文介绍 SpaceXAI的Grok Bot与OpenClaw拥有相同的编程能力，但其编程抽象层次不同。

Architecting memory and storage in the AI era

MIT Tech Review AI · 09/04 14:39

The era of AI inference has arrived. Imagine a healthcare system analyzing millions of data points in real time to accelerate life-saving medical research, or an intelligent assistant instantly resolving thousands of complex customer needs at once. These real-world breakthroughs rely on advanced inf

中文介绍 人工智能推理时代即将到来，设想一个医疗系统能够实时分析数百万数据点，以加速生命救助的医学研究。

Data from drones in Ukraine is fueling a new Wild West marketplace

MIT Tech Review AI · 09/04 05:25

Battlefields in Ukraine are littered with the remnants of drones, which are now firmly established as a critical weapon of modern warfare. But behind all that wreckage, there’s a new gold mine for the defense sector. The data drones generate will far outlast the wars in which they are used to fight,

中文介绍 乌克兰战场上散落着无人机残骸，而这些无人机生成的数据将为国防领域带来新的财富，超越当前的战争。

[AINews] GPT-6 Astra: OpenAI’s biggest LLM launch of all time

Latent Space · 09/04 01:18

new SOTA computer use and coding, 2.5x pricier per token, but WAY cheaper per task, less monitorable. overall, a very successful launch of OpenAI’s new frontier model class.

中文介绍 OpenAI推出的GPT-6 Astra是其历史上最重要的大语言模型，价格是每个token的2.5倍，但每个任务的成本显著降低。

GPT-6 Astra 🚀, Grok Bot Enterprise 🤖, Runway world model 🌎

TLDR AI · 09/03 20:00

GPT-6 Astra: an automated AI Engineer you can hire for <$6 an hour

Latent Space · 09/03 17:09

We spent 20B+ tokens of GPT-6 Astra to explore everything. Here’s our learnings.

中文介绍 通过使用超过200亿个GPT-6 Astra的token进行探索，得到了许多启示，雇用该AI工程师的费用低于每小时6美元。

Introducing WeatherNext 3, our most advanced and accurate global weather AI model

DeepMind Blog · 09/03 11:02

Daybreak for Frontline Defenders: $1B to protect essential services

OpenAI News · 09/03 09:15

OpenAI introduces Daybreak for Frontline Defenders. A $1 billion commitment expands access to frontier cyber AI, training, and support for essential services.

中文介绍 OpenAI推出“前线捍卫者日出”，承诺投资10亿美元，以扩大对边缘网络AI及关键服务的支持和培训。

NeoMME: an efficient Multimodal-native and Multilingual Encoder

Hugging Face Blog · 09/03 09:13

Playco cut manual fixes 50% prototyping games with GPT-6 Astra

OpenAI News · 09/03 08:00

Using GPT-6 Astra, Playco built three themed game prototypes from one grey box foundation and reported 50% fewer manual fixes than with the previous model.

中文介绍 Playco利用GPT-6 Astra基于单一基础模型构建了三个主题游戏原型，手工修正减少了50%。

Legora reviewed 41 documents in minutes with GPT-6 Astra

OpenAI News · 09/03 08:00

Legora used GPT-6 Astra to review 41 documents in minutes, find all four planted errors, and improve performance by nearly 40% in this financial-review workflow.

中文介绍 Legora使用GPT-6 Astra在数分钟内审核41份文件，发现所有四处于隐蔽错误，并提高了近40%的业绩。

GPT-6 Astra: A new generation of intelligence

OpenAI News · 09/03 07:00

Introducing GPT-6 Astra, our most intelligent and aligned model yet, with state-of-the-art capabilities across computer use, coding, cybersecurity, and science.

中文介绍 OpenAI推出GPT-6 Astra，具有先进的智能水平，覆盖计算使用、编码、网络安全和科学等领域。

[AINews] Muse Spark 1.3 matches GPT-5.6-Sol, confirming Meta Superintelligence as the newest Frontier Lab, >90% discount for training

Latent Space · 09/03 00:38

an epic comeback story for Meta

中文介绍 Meta的Superintelligence在新的前沿实验室中确认Muse Spark 1.3与GPT-5.6-Sol匹配，培训费用优惠超过90%。

Safety overview: GPT-6 Astra

OpenAI News · 09/02 20:00

GPT-6 Astra is our most capable broadly deployed model and our first to reach the Critical level of cybersecurity capability under our Preparedness Framework.

中文介绍 GPT-6 Astra是OpenAI最强大的广泛部署模型，首次达到其准备框架下的网络安全关键能力。

Fine-tuning a 350M Model for Better Structured Outputs in 100 GRPO Steps

Hugging Face Blog · 09/02 20:00