2026-10-26

Inherited, Not Invented: Tracing the Provenance of Wait in Reasoning Models

👍 15

10/22 16:03

Reinforcement learning (RL) is credited with teaching reasoning models to write {wait} before reconsidering a step, yet large RL accuracy gains do not require the word. This view can confuse what RL discovers with what a model inherits from its training data. We trace the word through open models' t

Different Tasks, Different Timesteps: Temporal Task-Aware Quantization for Diffusion Language Models

👍 23

10/22 08:00

Diffusion Language Models (DLMs) introduce unique quantization challenges due to their iterative denoising process. In this work, we reveal that quantization sensitivity is both taskand time-dependent, with different downstream tasks exhibiting distinct precision requirements across denoising steps.

BrickWright: Agentic Creation of Physically Feasible Brick Builds

👍 13

10/22 08:00

Frontier coding agents are becoming increasingly adept at 3D modeling. Yet, they still face challenges when building objects meant to be assembled in the physical world, often creating invalid assemblies with floating, colliding, or disconnected parts. We argue that providing agents with tools for a

FilmFlow: Agentic System for 3D Virtual Film Cinematography

👍 31

10/22 08:00

Complex camera motion is central to visual storytelling, yet translating cinematographic intent into controllable, executable camera trajectories remains challenging. We introduce FilmFlow, an agentic framework that converts natural-language instructions or reference videos into explicit camera traj

RIDAR: Safeguarding Research Integrity in Autonomous Research Agents via Automated Detection and Runtime Intervention

👍 16

10/22 08:00

Autonomous research agents are increasingly used to automate large portions of scientific research, yet they are prone to research integrity violations (RIVs): behaviors that compromise the validity, reproducibility, or faithfulness of the research process. We present RIDAR, a framework for automate

IQuest-Q1: Advancing Foundation Capabilities for Agentic CLI Systems

👍 98

10/22 08:00

We present IQuest-Q1, a sparse Mixture-of-Experts (MoE) agentic foundation model with 320B total parameters (15B activated). IQuest-Q1 is developed through a three-stage process: (1) stable and efficient pre-training, combining Stable SplitNorm, Muon, and muP with a two-stage curriculum whose second

What Makes a Good Unmasking Policy? A Controlled Study of Learned Unmasking in Diffusion Language Models

👍 12

10/22 08:00

Masked diffusion language models generate tokens in parallel, but their efficiency depends on which predictions are committed. Learned unmasking policies improve the speed--accuracy trade-off, yet differ in supervision, inputs, and capacity, obscuring which choices matter. We view supervised policie

F2G-Pose: Geometry-Aware Foundation Feature Lifting for Direct RGB-D Category-Level Object Pose Estimation

👍 13

10/22 08:00

This paper studies category-level RGB-D object pose estimation, which recovers an object's 3D rotation and translation without instance-specific CAD models or reference views. This task is highly ambiguous due to intra-class variations, symmetries, and partial observations. Current solutions struggl

AnyGameRenderer: Entity-Consistent Generative Rendering Across Games and an Entity-Level Benchmark

👍 19

10/22 08:00

AI coding agents can now build small playable games in minutes, but the results usually look like programmer art: gray blockouts, colored geometric primitives, and flat semantic blobs. We present AnyGameRenderer, a video-to-video generative renderer that turns such footage into high-quality visuals

Cola2: Co-Adaptive Hierarchical Language Modeling with Continuous Semantic Compression and Balanced Discrete Generation

👍 21

10/22 08:00

Hierarchical language generation separates continuous semantic modeling from discrete text realization, but its effectiveness depends on what the latent space compresses and how the discrete generator uses it. We propose Cola2, a framework that jointly studies continuous semantic compression, the ba

AgentVQA: A Unified Benchmark for Agentic Visual Understanding

👍 17

10/22 08:00

Vision-language models (VLMs) can perform a broad range of tasks across diverse settings, yet their performance in agentic contexts remains poorly understood. Although online evaluations in simulators and real-world environments are essential for measuring interactive agent performance, they are cos

Is Complex Agentic Memory Always Needed? Naive RAG Is Stronger Than We Thought

👍 23

10/22 08:00

LLM memory is widely used in long-term conversation understanding. Agentic memory systems often use mutation, structural organization, learned retrieval strategies, or other complex operations. Is this complexity always necessary? We study Raw-RAG, a naive RAG baseline using dense retrieval and rera

Skeleton-Grounded Latent Action Model for Robot World Models

👍 7

10/22 08:00

Robot world models simulate how environments evolve in response to robot actions. Latent action models enable learning dynamics from videos by encoding visual transitions as proxy actions. However, reconstruction-driven latent actions mix robot motion with object responses and other scene changes, c

Beyond Watching and Searching: Benchmarking Agents for Video-centered Deep Research Reports

👍 25

10/22 08:00

Deep research agents synthesize open-web evidence into long reports, but current benchmarks do not evaluate research centered on a source video. We introduce Video-Centered Deep Research Report Generation. Given a video and user request, an agent identifies research questions, retrieves external mul

SCOPE: Compiling Natural-Language Policies into Neurosymbolic Predicate Programs

👍 15

10/22 08:00

Policy enforcement for LLM agents has shifted from single-turn content moderation to compound, application-specific policies evaluated over multi-turn transcripts and tool-call trajectories. Fine-tuned guardrail models are efficient but degrade on compound policies, and while LLM-as-a-judge approach

TARSphere: A Robot-Oriented Full-Stack Framework for Long-Horizon Embodied Agents in Evolving Physical and Social Environments

👍 8

10/21 08:00

Robots must act under physical and social constraints while adapting to changes in their environment and human needs. Existing embodied-agent systems often focus on specified goals or abstract execution into symbolic actions, limiting evaluation in an evolving world. We introduce TARSphere, a robot-

Agent Mosaic: Recognizing Confidentiality Leaks in Agentic LLM Systems

👍 23

10/21 08:00

Modern LLM agents can retrieve information from accessible documents, and then disclose a sensitive conclusion assembled from them. This mosaic effect creates a critical risk for agents operating over confidential data. We quantify its realization with 30 synthetic cybersecurity, medical, and milita

Toward Recursive Self-Improvement in Physical AI

👍 59

10/21 08:00

Recursive self-improvement (RSI) describes a system that improves not only its behavior but also the process by which it improves. Recent digital agents test this idea by rewriting their own code and checking each version with executable tests. Bringing these mechanisms into the physical world requi

RLSplat: Resource-Linear Budget Control for Feed-forward 3D Gaussian Splatting

👍 23

10/20 08:00

Feed-forward 3D Gaussian Splatting (FF3DGS) reconstructs 3D scenes from images in a single forward pass, making it well-suited to applications that must operate within limited resource budgets. To further improve its practicality, recent budget-controllable FF3DGS methods allow the number of Gaussia

Diffusion Language Models Change Speed Across Tasks and Languages

👍 10

10/19 13:47

Diffusion language models can write several tokens in one forward pass, and they decide during generation how many passes to spend. Two outputs of the same token length can therefore vary a lot in cost. To understand the underlying factors of this cost variation, we analyze two diffusion models acro

Who Takes the Lead? Model Ability and Task-Conditioned Collaboration with Humans

👍 10

10/19 08:00

As language-model agents grow more capable, they can carry more of a task without asking anyone for help. Falling reliance on a human partner is easy to read as a failure of collaboration when it may instead be a correct division of labor, and the rate of human involvement alone cannot tell the two

RouteWeaver: Weaving Mode Selection and Execution into Unified LLM Routing

👍 25

10/19 08:00

LLM routing dynamically selects and coordinates heterogeneous language models across diverse tasks. However, existing approaches are largely built around a predefined execution paradigm, such as single-round, multi-round, or agentic routing, which fixes how model calls are organized and limits adapt

Heartbeat Thinking: Speculative Reasoning and Tool Execution While Listening in Realtime Voice Agents

👍 24

10/19 08:00

Tool-using voice agents often leave users waiting after speech ends, even when actionable information arrives earlier in the request. Speculative approaches begin work during speech, but chunk-bounded or interrupt-on-update generation can truncate a tool call before its arguments are complete. This

Deep Search in Real-World Web Environments

👍 164

10/19 08:00

Despite substantial progress in Deep Search agents, their ability to conduct end-to-end research in real-world open Web environments remains insufficiently evaluated. Unlike static retrieval settings, information on the real-world Web is distributed across diverse sources, modalities, and heterogene

Deep Fusion Memory: A Chunk-Level Parametric Memory for Large Language Models

👍 50

10/18 08:00

Large language models need effective ways to acquire and use external knowledge beyond what they retain from pretraining. Continued pretraining risks catastrophic forgetting, while retrieval at inference time can add substantial overhead. External parametric memory avoids online retrieval, but const

Bridging Exploration and Persona: Information-Directed Dueling for Test-Time Agent Personalization

👍 7

10/17 08:00

Effective test-time personalization requires LLM agents to adapt to query-dependent user preferences from lightweight interaction. We study persona selection with binary comparisons between candidate responses as the online signal. We present ID-TAP, a scalable, IDS-inspired neural dueling-bandit fr

IF-MAP: Influence-Function-Inspired Agent Prioritization in Multi-Agent Systems

👍 8

10/17 08:00

Improving every agent in an LLM-based multi-agent system can be prohibitively costly, making it important to identify which agent should be prioritized to achieve the greatest system-level gain with limited resources. Failure attribution can inform this decision, but an agent's responsibility for fa

EvalResearchBench: Can AI Agents Build Their Own Feedback for Recursive Self-Improvement?

👍 22

10/16 08:00

Recursive self-improvement (RSI) relies on evaluation feedback to assess progress and guide further research, yet repeatedly running complex benchmarks is costly and slows iteration. Human experts reduce this cost by selecting benchmark subsets or designing compact suites. We ask whether AI agents c

The LLM-Attacker: A Unified Red-Team Framework for Privacy Evaluation of Split and Distributed LLM Systems

👍 10

10/15 08:00

Split language models keep input text on a client but send intermediate representations to a server. Attacks can reconstruct text from these messages. Comparing privacy evaluations is difficult when attackers see different messages, know different model weights or use different datasets and recovery

MolLean: A Process-Verifiable Language for Chemical Reasoning

👍 15

10/14 08:00

Chemical chain-of-thought can contain hallucinated structural and property claims. We thereby introduce MolLean, a process-verifiable language that grounds chemical reasoning in an explicit molecular state. A layered Lean-based kernel checks state transitions and assertions over molecular readouts,

how to build and run an army of AI UGC accounts

@tibo_maker · 215.7K 粉丝 · 56.5K 阅 · 510 赞 · 30 转

10/23 22:06

in this article I'll walk you through all 9 steps of the setup, with the 12 prompts I use, ready to copy, in the order you'll need them. you've probably scrolled past a few of them this week. a girl

中文介绍 tibo_maker分享如何构建并运行一支AIUGC账号团队，详细介绍了9个步骤，并提供12个可直接复制的提示词。

The five components of building a "company harness"

@hwchase17 · 136.6K 粉丝 · 39.2K 阅 · 521 赞 · 60 转

10/25 01:08

How do you build a company harness? We've built internal agents at LangChain across marketing, sales, and engineering, and they have completely changed how those teams work. They weren't easy to get

中文介绍 hwchase17解析构建“公司 harness”的五个组成部分，以LangChain内部在营销、销售和工程中构建的内部代理为例。

3 markdown architecture patterns that boost agentic productivity

@GoogleCloudTech · 1.3M 粉丝 · 38.7K 阅 · 502 赞 · 68 转

10/23 09:33

There is a running joke in modern software engineering that the hottest programming language isn't Rust, Zig, or Python: it’s Markdown. As coding agents become entrenched in developer workflows, many

中文介绍 GoogleCloudTech介绍三种Markdown架构模式，旨在提升编码代理的工作效率。

How I work with Agents

@thorstenball · 57.2K 粉丝 · 35.8K 阅 · 831 赞 · 59 转

10/25 23:58

... as of October 2026. Use the latest and greatest models. Everything else is potentially a waste of time. If a model is wrong 10% of the time, you don't know which 10%, so you end up making multiple

中文介绍 thorstenball分享如何与代理协同工作，强调使用最新的模型，认为错误的10%未知导致多次尝试是浪费。

[AINews] Kimi 3.1 Pro 104BA-2.8T: Fable-5.1-Class Open Weights narrows Closed-AI lead to 1 month... and Pro Highspeed API with 2-3× speed

Latent Space · 10/24 15:18

We have Fable at home... or so we are being told by paid shills

中文介绍 Kimi 3.1 Pro 104BA-2.8T推出，缩小与Closed-AI的差距，并推出2-3倍速度的Pro Highspeed API。

Rapidly scaling online storage to serve over 1 billion ChatGPT users (Part II)

OpenAI News · 10/24 01:00

Learn how OpenAI scaled Habitat with reliable multi-tenancy, global caching, and Azure Cosmos DB to deliver 12× faster reads for ChatGPT.

中文介绍 OpenAI通过可靠的多租户、全局缓存和Azure Cosmos DB，将Habitat的读取速度提高12倍。

LitPivot: Developing research ideas through dynamic contextualization within the literature landscape

Hugging Face Blog · 10/24 00:15

中文介绍 Hugging Face博客介绍通过文献环境中的动态语境化来发展研究想法。

AI Is Getting Good at Math. Here’s Why.

Latent Space · 10/23 23:19

AI models are solving more mathematical problems. A pattern is emerging in their successes that could help us predict which problems they’ll solve next.

中文介绍 AI模型在解决数学问题方面取得进展，出现成功模式，有助于预测未来解决的问题。

[AINews] Black October

Latent Space · 10/23 18:09

the Cyber onslaught continues.

中文介绍 网络攻击持续进行中。

Securing our frontier model training environment

OpenAI News · 10/23 18:00

OpenAI explains how we strengthened containment for frontier model training and evaluation, including offline environments and high-compute red-teaming.

中文介绍 OpenAI加强了对前沿模型训练和评估的隔离，包括离线和计算强化。

OpenAI and Poland’s NASK partner to advance AI safety and security

OpenAI News · 10/23 08:00

OpenAI and Poland’s NASK are partnering on AI safety, security, and technical research to strengthen safeguards and support a more resilient digital ecosystem.

中文介绍 OpenAI与波兰NASK合作，推进AI安全和网络安全，支持更坚韧的数字生态系统。

Does Your Model Run on AWS Trainium? Look It Up

Hugging Face Blog · 10/23 08:00

中文介绍 Hugging Face博客介绍AWS Trainium上模型的运行情况。

OpenAI nears $1.25T 💰,  ☁️, 1Password for Agents 🔑, Gemini 3.5 Pro postmortem 🤖

TLDR AI · 10/23 08:00

中文介绍 OpenAI估值接近1250亿美元，推出1Password for Agents，Gemini 3.5 Pro分析。

Turning Database Grading into RL Rewards: Post-Training Agents with ThinkingBox

Hugging Face Blog · 10/23 06:24

中文介绍 Hugging Face博客介绍如何使用ThinkingBox将数据库评分转化为强化学习奖励。

Designing an AI serving harness built to last

MIT Tech Review AI · 10/23 05:17

The next competitive advantage in enterprise AI has less to do with which model you choose than with what you build around it. That’s an important shift. For the past several years, much of the AI conversation has centered on model capability: bigger models, better reasoning, larger context windows,

中文介绍 企业AI的下一个竞争优势在于构建于模型周围的东西，而非模型本身。

Build more responsive apps with the Decisions API

OpenAI News · 10/23 03:00

Fast, focused, low-cost decisions for any application.

中文介绍 OpenAI推出Decisions API，提供快速、专注、低成本的决策。

The Foundations of Scalable AI

MIT Tech Review AI · 10/23 02:48

A scalable AI system requires a strong foundation. Data, engineering, platforms, operating models, and governance must converge to work as one. This session from MIT Technology Review’s EmTech Future conference explores what it takes to build that foundation in a highly regulated environment and how

中文介绍 可扩展的AI系统需要坚实的基础，数据、工程、平台、运营模式和治理必须协同工作。

Why AI Compute Comes in Rectangles — Erik Bernhardsson, Modal CEO

Latent Space · 10/23 01:40

Elastic GPUs, sandbox growth, and the economics of the agent cloud

中文介绍 Modal CEO Erik Bernhardsson探讨AI计算为何以矩形出现，以及沙盒增长和代理云的经济。

Beyond tree cover: mapping the world’s forest types

DeepMind Blog · 10/22 22:30

A global map categorizing every 10-meter patch of land on Earth into distinct forest types.

中文介绍 DeepMind发布全球地图，将地球每10米土地划分为不同的森林类型。