2026-10-14

Evaluating LLM Creativity through Long-Tail Performance

👍 3

10/09 08:00

Creativity benchmarks for large language models (LLMs) vary widely in task and metric design. This raises the question of whether they capture a common notion of creativity. We propose that this shared target can be formalized as task-specific performance in the long tail of a reference input-output

Branching On-Policy Distillation

👍 5

10/09 08:00

On-policy distillation (OPD) provides teacher supervision on student-generated trajectories, but a single trajectory realizes only one continuation at each decision, limiting supervision coverage under a finite sampling budget. We propose BranchOPD (Branching On-Policy Distillation), which selects i

Learning More by Training Less: Uncertainty-Guided Progressive Contraction for Efficient Reasoning Post-Training

👍 7

10/09 08:00

Large language models have achieved remarkable advances in complex reasoning. However, scaling reasoning post-training with ever-larger corpora can substantially increase training cost without proportional performance gains. In this study, we therefore ask: how far can reasoning supervision be compr

AgentSWE: Can Coding Agents Build the Agent You Actually Want?

👍 201

10/09 08:00

As the production of software passes into the hands of software itself, coding agents are increasingly commissioned to build agents fitted to their users' own workflows. This delegation carries a quiet assumption, that the delivered agent will faithfully serve the requirement it was built for, rathe

Scaling Agentic Data for Long-Horizon Terminal Intelligence

👍 133

10/09 08:00

Terminal agents are moving beyond local operations toward completing real-world professional tasks. Short trajectories can demonstrate commands and tool calls, but they do not fully show how agents accumulate evidence through exploration, use it to inform judgments, continually adjust strategies bas

SpatialASM: Internalizing Active Mentalization for Spatial Reasoning

👍 16

10/09 08:00

When solving complex spatial reasoning tasks, ordinary humans naturally revisit the scenes from task-relevant viewpoints and mentally reconstruct the spatial layouts from those viewpoints. But can Large Vision-Language Models (LVLMs) proactively think in the same way? We term this ability Active Spa

Multi-Stream Attention Residuals

👍 34

10/09 08:00

Attention Residuals (AttnRes) enables content-dependent retrieval over network depth, which can be viewed as single-stream softmax attention along the depth axis. Inspired by the success of multi-head attention along the sequence axis, we introduce Multi-Stream Attention Residuals (MSAR), an extensi

GATOR: Generative and Agentic 3D Object Reconstruction From Casual Images

👍 1

10/08 12:13

Reconstructing complete, scene-aligned 3D objects from casual images requires integrating sparse, uncertain observations and inferring surfaces hidden by occlusions. We present GATOR, a generative and agentic framework that recovers textured object assets and their scene-relative pose from one or mo

UNITAS: A 3D-Native World Action Model for Embodied Manipulation

👍 1

10/08 08:00

World action models (WAMs) aim to answer a coupled physical question: given a task instruction, what motion should the robot execute, and how will that motion change the surrounding world? Most existing WAMs build on pretrained video generators and represent world evolution through images or visual

Thinking Inertia: LLMs Keep Thinking When Told Not To

👍 215

10/08 08:00

Large Language Models (LLMs) increasingly ship with explicit "thinking modes", yet their counterpart, "no-thinking", has received far less attention. We study LLMs' no-thinking behavior along two axes. a. How to measure no-thinking? Prior work typically defines no-thinking through proxies such as a

Humanity's Sixth Sense: Benchmarking Intuitive Visual Reasoning in Multimodal Models

👍 21

10/08 08:00

Humans perceive far more in a scene than what is explicitly depicted: a single glance captures past causes and future trajectories; a quick peek determines if a vehicle can fit between two parked cars; a few seconds of video reveals who holds authority in a room; and a fleeting clip highlights subtl

Mental-Models for Multi-Agent Systems

👍 4

10/08 08:00

Large foundation models have accelerated progress toward general-purpose agents that interact with humans and other agents through language and multimodal signals. However, robust multi-agent decision-making requires reasoning about what other agents know, intend, and are likely to do under partial

Clinician use of language models diverges from how the models are evaluated

👍 3

10/08 08:00

Large language model (LLM) assistants are being deployed to clinicians across health systems, and judgments about their readiness rest largely on benchmark scores, most of them derived from examination questions or curated cases. A benchmark predicts performance in deployment only to the extent that

One Model to Loop Them All: Elastic Latent Recurrence for Language Models

👍 7

10/08 08:00

A latent recurrent language model applies a shared block of layers repeatedly, so one checkpoint can trade inference compute for accuracy by changing the loop count. At one loop, however, such models typically fall well short of a non-recurrent model of the same size, a one-loop tax that makes a sep

Self-Preservation in Language Models

👍 3

10/08 08:00

Continual learning (CL) seeks to adapt pretrained models while preserving prior capabilities. Yet the tasks supplied for preservation cover only part of what a model knows, while forgetting can extend beyond them. Recent work shows that language models can supplement this limited evidence by generat

VIS-Ground: Video Interactive Storytelling with Contextual Grounding

👍 78

10/07 08:00

Video interactive storytelling enables viewers to actively steer how a video unfolds. However, once we allow viewers to intervene during generation, a new challenge arises: The viewer's request can have latent dependencies on both the grounding source and the current rendered video state. These depe

Systematic Multi-Agent Vision-and-Language Navigation: Formulation, Benchmark, and Method

👍 10

10/07 08:00

Vision-and-Language Navigation (VLN) has largely focused on a single agent following a single instruction, yet many real-world applications require teams of robots to tackle tasks beyond the capabilities of any individual agent. We present Systematic Multi-Agent Vision-and-Language Navigation, provi

Coding-Agent Benchmarks Should Match Their Users' Task Flows

👍 11

10/07 08:00

The evaluation of coding agents generally strives to be as realistic as possible. In our study, we collect 4,782 agent sessions of real software engineers in JetBrains IDEs, which we call Production Sessions. Since our subject is interactive agents, we study the sessions with at least three user mes

SPLATIFY: Reproduce, Discover, Innovate! From Papers and Ideas to Trainable 3DGS Code

👍 1

10/06 08:00

The rapid growth of 3D Gaussian Splatting (3DGS) research demands significant effort to reimplement papers before building on them. We introduce SPLATIFY, a multi-agent framework that converts 3DGS papers into trainable gsplat-based implementations, where generic paper-to-code methods and frontier m

Visual Grounding Safety in Vision-Language Models

👍 2

10/05 08:00

Vision-language models (VLMs) are increasingly trained to generate structured outputs like points and bounding boxes that downstream interfaces, agents, and robots can act on, yet safety alignment of this output channel has not been systematically analyzed. We study visual grounding safety by repurp

Distributionally Robust Mixture-of-Experts Training

👍 6

10/05 08:00

Mixture-of-Experts (MoE) transformers scale capacity by activating only a few experts per token, but this sparsity creates a hidden reliability problem: when routing is imperfect, load-balanced models may send tokens to experts that are insufficiently trained for the assigned inputs. We propose Dist

HyperThink: Text-to-Parameter Hypernetworks for Efficient Reasoning

👍 0

10/02 08:00

Long-form thinking traces can substantially improve the multi-step reasoning performance of large language models (LLMs), but they introduce high inference-time overhead, with latency dominated by sequential decoding. We propose HyperThink, a text-to-parameter approach that amortizes this reasoning

Prospective Hindsight: Self-Calibrating Reinforcement Learning via Prediction-Reality Gaps

👍 3

10/02 08:00

Reinforcement learning for long-horizon agents relies on purely retrospective training signals: credit is assigned only after observing environmental consequences, leaving the agent's belief at action time invisible to the gradient. We introduce Prospective Hindsight (PH), a self-calibrating trainin

T3lescope: Arbitrary-Resolution High-Fidelity Generative Surface Reconstruction from Images

👍 10

10/02 08:00

We reconstruct high-fidelity 3D scene meshes from posed multi-view images without per-scene optimization, across scales ranging from single objects to large outdoor scenes. Per-scene optimization methods lack the learned 3D prior needed when observations are sparse or surfaces are glossy or transpar

Initialization Improves LLM-Driven Discovery

👍 4

09/30 08:00

Large Language Models (LLMs) have been used for novel discovery of algorithms, theorems, drugs, and other tasks through the use of harnesses that prompt an LLM to iteratively optimize an objective. In this work, we study the relationship between the population of previous iterates and eventual disco

GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed

👍 73

09/30 08:00

Autoregressive (AR) grounding models serialize spatial predictions, introducing sequential latency and imposing a causal order on output tokens. We view grounding as visual evidence extraction: objects, locations, and spatial relations are jointly constrained by the image and query, yet their depend

GroundingPI: A Grounding Foundation Model towards Physical Intelligence with Visual Primitives

👍 247

09/30 08:00

Precise grounding matters. It specifies which object is the target and where that object is, even in clutter and for tiny objects, and it has to be fast enough for closed-loop control. Yet vision-language-action (VLA) and world-action models (WAMs) take perception from general-purpose vision-languag

PR-OPD: Privileged Representation On-policy Self-Distillation for Agentic Reinforcement Learning

👍 20

09/29 08:00

Language-model agents are usually trained by reinforcement learning from one reward per episode, and privileged self-distillation enriches it by letting the same policy, given a skill, teach its skill-free self through token probabilities. However, we identify two phenomena that question this channe

CrossTimeEdit: A Decade-Spanning Cross-View Dataset and Reward-Guided Editing for Historical Street-View Generation

👍 20

09/29 08:00

Historical street-view imagery records urban evolution, but uneven coverage leaves substantial gaps in historical records. Generating plausible past appearances requires restoring changed structures while preserving persistent scene content. We construct VIGOR-his, a decade-spanning cross-view datas

Relevance Does Not Imply Applicability: Experience Activation for Personal GUI Agents

👍 30

09/27 08:00

Personal Graphical User Interface (GUI) agents rely on interaction history to infer what a user wants from ambiguous instructions and to anticipate recurring routines. Existing approaches retrieve task-relevant history and append it to the policy's context, implicitly assuming that experience releva

How Grok Bot Changed the Way I Work

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

10/13 13:20

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 博主介绍如何利用Grok Bot改变工作方式，分享使用Claude Code、Codex和Cursor等AI编程助手的经验。

Introducing Glass: Learn Things Once.

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

10/13 22:00

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Ramp的快速迭代，强调为员工提供最佳AI模型和工具的重要性。

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

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

10/13 00:52

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 博主分享35个成人AI图像和视频生成平台，涵盖成人AI图像、视频创作、动漫等。

🥇Top AI Papers of the Week

@dair_ai · 134.1K 粉丝 · 54.0K 阅 · 538 赞 · 56 转

10/11 21:10

1. Harness Learning An agent is defined by its model and its harness, the program that organizes model calls, tool use, and information flow. Researchers at CMU train a proposer model to revise the

中文介绍 博主推荐本周顶级AI论文，包括CMU研究人员训练的提案模型。

Game of skill

@kirat_tw · 124.9K 粉丝 · 22.0K 阅 · 503 赞 · 26 转

10/11 21:56

For most of the last two decades, coding was the filter. If you could write good software, you were “smart” in the way the market cared about. Elite universities, top firms and startups all used it as

中文介绍 博主讨论技能竞赛中的编程角色，以及如何成为市场上认可的“聪明人”。

Gemini 4 Argon: Built to solve your hardest problems

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

10/14 01:33

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 GoogleAIStudio发布Gemini 4 Argon，适用于开发者、企业，以及Google AI Ultra订阅者。

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

Latent Space · 10/14 06:25

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首席科学家押注世界动作模型与物理AI新前沿，推出首个机器人舱Dyna-2。

Developing Gemini 4 Argon responsibly and safely

DeepMind Blog · 10/14 03:43

中文介绍 DeepMind推出Gemini 4 Argon，负责负责和安全的开发。

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

Hugging Face Blog · 10/14 01:05

中文介绍 Hugging Face推出LFM2.5-24B-A2B，在设备上快速运行私有智能体。

Gemini 4 Argon for Robotics

DeepMind Blog · 10/14 00:01

中文介绍 DeepMind发布Gemini 4 Argon，用于机器人技术。

Gemini 4 Argon: Built to solve your hardest problems

DeepMind Blog · 10/14 00:01

中文介绍 DeepMind推出Gemini 4 Argon，旨在解决最复杂的问题。

Higgsfield takes AI video from 1 minute to over 10

OpenAI News · 10/13 20:00

Higgsfield brings creators longer videos with GPT-6 Astra, precise visuals with GPT-Image-2.5, and a path to voice filmmaking with GPT-Live.

中文介绍 OpenAI的Higgsfield使用GPT-6 Astra和GPT-Image-2.5提供更长的视频和精确的视觉效果，并探索语音电影制作。

Open models have real benefits. Do they outweigh the risks?

MIT Tech Review AI · 10/13 18:00

MIT Technology Review Explains: Let our writers untangle the complex, messy world of science and technology to help you understand what’s coming next. You can read more from the series here. Late last month, Anthropic released a report arguing that the model GLM-5.3 poses a major cyberattack risk. T

中文介绍 MIT Technology Review解释开放模型的实际益处和潜在风险。

[AINews] Meta Muse Wonder & Wonder Full: MSL’s Claude 5.5-tier Frontier Model

Latent Space · 10/13 17:37

MSL’s emergence as a frontier lab is finally complete.

中文介绍 Meta Muse Wonder & Wonder全系列：MSL的Claude 5.5层前沿模型正式亮相。

AI labs are buying data from dead startups to train their models

MIT Tech Review AI · 10/13 17:00

When Keelan Miskell began winding down his crypto startup this year, he didn’t think its data was worth much: the users were anonymous, the company just a few years old. But he’s since turned the firm’s digital archives into real money. He’s netted roughly $25,000 for his code base and a still-pendi

中文介绍 AI实验室从倒闭的初创公司购买数据来训练模型，如Keelan Miskell的公司代码档案。

Muse Wonder ✨, Maia 300 ⚡, AI self-replication 🧬

TLDR AI · 10/13 08:00

中文介绍 Muse Wonder、Maia 300和AI自我复制技术介绍。

How Anthropic’s Dynamic Workflows Coordinate Claude Agents

Latent Space · 10/13 02:18

At AI Engineer New York, Anthropic drew on lessons from Unix and Linux to explain how agents can tackle bigger projects without overwhelming developers.

中文介绍 Anthropic的动态工作流程协调Claude智能体，从Unix和Linux经验中汲取教训。

Gemini 4 Carbon 🧠, Nvidia weighs Reflection acquihire 💰, decision model race 🤖

TLDR AI · 10/12 08:00

中文介绍 Gemini 4 Carbon、Nvidia收购Reflection和决策模型竞赛等AI新闻。

[Subscriber Exclusive] NYC Subscriber Meetups!

Latent Space · 10/11 06:07

Tomorrow and Tuesday. If you see this you’re in!

中文介绍 Latent Space举办纽约订阅者聚会。

Building AI for Reliable Execution: Lessons From Industrial Robotics

Latent Space · 10/10 22:04

Inside Standard Bots’ AI stack, pretrained models learn factory tasks from demonstrations and improve through corrections from real deployments.

中文介绍 Standard Bots从工业机器人经验中学习，建立可靠执行的AI系统。

[AINews] TypeSafe/Jev at >$100M ARR, $7.5B valuation 3 weeks after launch

Latent Space · 10/10 14:45

Wow.

中文介绍 TypeSafe/Jev在发布3周后达到超过1000万美元的年度经常性收入和75亿美元的估值。