2026-08-27

When "Must" Becomes "Maybe": Constraint Weakening in LLM Agent Workflows

👍 3

08/25 08:00

Large language model (LLM) agents coordinate complex tasks through multi-role and multi-stage workflows. Upstream state is repeatedly transformed into intermediate language artifacts, such as summaries, plans, tickets, memories, and handoff notes, from which downstream components act. For action-con

MoTE: Mixture of Task Experts for Multi-Task Video Understanding

👍 3

08/25 08:00

Procedural video-language models must solve heterogeneous tasks from the same visual evidence, including action recognition, forecasting, and procedure prediction. Dense transformer decoders share the same feed-forward networks across tasks, which can entangle task behavior and make controlled capab

Latent Action as Intention Enables Efficient Future Imagination for World Action Models

👍 0

08/25 08:00

World action models (WAMs) improve robot control by modeling how observations evolve, but generating future observations at test time incurs substantial latency. Fast-WAM removes this process for efficiency; however, our matched implementations show lower generalization for Fast-WAM than for future-

Game2World Engine: Unlocking In-the-Wild Gameplay Videos for World Model Training

👍 8

08/25 08:00

Video games provide a scalable source of training data for video world models, offering diverse environments, complex interactions, and abundant in-the-wild gameplay videos. However, raw gameplay footage entangles the game world with screen-space interfaces, introducing game-specific biases and irre

From Seeing to Acting: Smart Glasses as First-Person Intelligence Platforms

👍 8

08/25 08:00

Smart glasses are evolving from capture and display accessories into first-person intelligence platforms that connect human perception, persistent context, and digital or physical action. Their on-body viewpoint aligns with the wearer's vision, audition, motion, and hand-object interaction, but must

WeMM-Embedding: WeChat Multi-Modal Embedding Technical Report

👍 54

08/25 08:00

Universal multimodal embeddings are becoming a core component of modern AI systems, enabling heterogeneous content to be represented in a shared space for applications such as retrieval, recommendation, classification, and agentic systems. In this report, we present WeMM-Embedding, a family of unive

Best Practice Critic Optimization

👍 12

08/25 08:00

Group-based reinforcement learning methods such as GRPO for large language models avoid training a critic by sampling multiple responses for each prompt. A reliable critic could instead estimate token-level advantages from one response, but standard critic-based training recipes are often unstable.

On-policy Distillation with Verifiable Reward

👍 12

08/25 08:00

Reinforcement Learning with Verifiable Rewards (RLVR) and on-policy distillation (OPD) have become two widely adopted paradigms for post-training large language models. However, RLVR suffers from sparse task-level feedback, while OPD provides dense token-level guidance but ignores trajectory correct

On-Policy Self-Distillation in Diffusion Models

👍 48

08/25 08:00

Reinforcement learning can align diffusion models with human preferences and task-specific objectives, but endpoint rewards do not specify how an intermediate denoising prediction should change. We introduce DiffusionOPSD as an on-policy self-distillation framework that converts image-level reward g

Recursive Experiential-Working Memory Evolution for Long-Horizon Agent Harnesses

👍 20

08/25 08:00

Recursive self-improvement (RSI) remains hard in long-horizon tasks, where growing histories obscure the task state and misalign skill invocation. We introduce Recuris, a recursive Experiential-Working Memory architecture for long-horizon agent harnesses, in which Working Memory tracks task progress

Meta^n: Recursive Self-Improvement through Emergent Depth

👍 8

08/25 08:00

Self-improving LLM agents refine answers, not the process that produces those answers. Systems that add a meta-level hold that level fixed, and those that edit themselves must leave part of their own editing machinery untouched to stay stable, capping the meta-depth they realize at roughly two. We p

CAFE: Self-Improving Search Agents Need Co-Evolving Feedback

👍 4

08/25 08:00

Outcome-supervised search agents learn when and how to retrieve evidence, but terminal rewards neither localize intermediate errors nor redirect an ongoing trajectory before those errors compound. Treating corrective feedback as a learned in-trajectory intervention couples the two roles: the agent m

Autonomous Mathematical Discovery in an Open-World Multi-Agent Environment

👍 2

08/24 08:00

We study autonomous mathematical discovery in the Station, an open-world multi-agent environment in which AI agents from different model families pursue a shared research goal without a central coordinator or scripted pipeline. Agents choose their own research directions, conduct experiments, collab

AgentRoom: Concurrent Multi-Agent Coding in a CRDT-Backed Shared Workspace

👍 4

08/24 08:00

Concurrent multi-agent coding promises division of labor across modules, robustness through redundancy, and parallel exploration at the natural granularity of multi-file projects. Realtime collaborative editing protocols solve this coordination problem for human teams via Conflict-free Replicated Da

Automata from Agent Traces: Failure and Next-Step Prediction

👍 4

08/24 08:00

LLM-based agents execute multi-step tasks, but their behavioral structure remains opaque: long unstructured traces resist the safety auditing and runtime monitoring that deployment requires. Existing approaches operate per-trace or success-only, so they miss the cross-run topology that links next-st

CyberFactory: Scaling Cyber Security Capabilities with Instances from the Wild

👍 28

08/24 08:00

As large language models (LLMs) continue to advance in coding capabilities, their potential in cybersecurity has drawn increasing research attention, with closed-source LLMs (e.g., Mythos) delivering advanced cybersecurity capabilities. However, existing open-source efforts remain limited: frontier

AutoSaddler: Automatic Harness Optimization with Durable Updates from Agent Execution Traces

👍 49

08/24 08:00

LLM agents remain unreliable on long-horizon tasks, where small local failures can compound over extended interactions and lead to overall task failure. Although external harnesses can substantially improve robustness, harness design remains a manual and expensive process that requires searching ove

The Laws of Context Allocation: Causal Measurement and Closed-Loop Orchestration in Generative Search

👍 4

08/24 08:00

As Retrieval-Augmented Generation (RAG) shifts toward diverse portfolio generation, it is stymied by two critical bottlenecks: flawed measurement of evidence utilization, and suboptimal context budget allocation. We resolve both sequentially. To resolve measurement, we expose a pervasive ``diagnos

LongWoF-Bench: Evaluating EvoMap Genes for Verifiable Long-Workflow Tasks

👍 6

08/24 08:00

Large language models are increasingly expected to execute complex workflows whose success depends on maintaining interdependent constraints and producing artifacts that satisfy strict end-to-end verification. Yet successful execution experience is typically lost after a single run, forcing subseque

Length-Adaptive Decoding for Masked Diffusion Machine Translation

👍 4

08/23 08:00

Machine translation tests masked diffusion language models (dLLMs) because every source token must be rendered faithfully, while fixed canvas decoding must choose target length before denoising. Existing masked diffusion decoding work mainly studies token unmasking order, leaving this length decisio

What AstroPT knows about galaxies, and what that can teach us about LLMs

👍 4

08/23 08:00

Interpretability research increasingly asks when concepts emerge during training and whether linear probes recover real structure, but in language models these claims are hard to validate because language offers little ground-truth ordering of concepts or relationships among them. We propose the use

ClawProBench: Trace-Aware Evaluation of AI Agents with Runtime Coverage and Frozen Workplace-Style Holdouts

👍 4

08/23 08:00

Agent benchmarks often evaluate only final answers even when agents run on stateful runtimes. We argue this under-specifies what is being evaluated: the proper unit is a declared model-plus-runtime configuration whose failures can occur in evidence acquisition, runtime routing, safety boundaries, or

AutoResearch: Insight In, Hallucination Out

👍 8

08/23 08:00

Autonomous research systems are increasingly capable of executing long research workflows, yet automation alone does not ensure that the resulting process remains scientifically grounded. We introduce AutoResearch, a two-stage system that connects Idea Generation with Idea Execution to address both

SecOPD: Mitigating Adaptive Prompt Injections by On-Policy Distillation

👍 36

08/21 08:00

Prompt injection is listed as the \#1 threat to AI agents. When an agent accesses external data from websites, files, or emails, an attacker may inject a prompt into the data, saying, "Ignore all prior instructions and perform <an attacker's task>." To prevent arbitrary manipulation of agents, defen

Quantization-Aware Healing: A Practical Recipe for Recovering Compressed, 4-Bit LLMs

👍 9

08/21 08:00

Serving large language models cheaply increasingly means shipping models that are both structurally compressed to a fraction of their parameters and quantized to 4 bits. Together these steps degrade reasoning, mathematics, coding, and long-context behavior enough to require a recovery, or healing, s

Annotations as Rollouts: Efficient and Scalable Reinforcement Learning for Video MLLMs

👍 88

08/20 08:00

Multimodal large language models (MLLMs) have become a prevailing paradigm for unified video perception. However, post-training on large multi-task datasets remains challenging, as existing reinforcement learning methods sample on-policy groups with few high-quality rollouts even with costly chain-o

Tomatoes, Potatoes, and Onions: Questioning the Need for Faces in Face Presentation Attack Detection

👍 3

08/20 08:00

Face presentation attack detection (PAD) is traditionally formulated as a face-specific problem, although many of the visual artifacts introduced by print, replay, and recapture processes are not inherently tied to facial appearance. In this work, we investigate whether transferable PAD representati

GigaBrain-0.7: Scaling Embodied Foundation Models to Emergent Capabilities with a Three-System Architecture

👍 90

08/16 08:00

Vision-language-action (VLA) models have become a dominant paradigm for generalist embodied agents, demonstrating strong complex and long-horizon task completion in structured settings. Yet it remains an open question whether current VLA systems can benefit from more effective architectural design,

LongRCA Bench: Diagnosing Responsible Roles and Root Causes in Long-Horizon Agent Failures

👍 14

08/15 08:00

When a long-horizon agent execution fails, outcome-level evaluation reveals the unsuccessful result but not where the decisive error entered the trajectory. Developers must then inspect the full execution to identify the responsible role and localize the earliest decisive root-cause step. Existing f

DREAM Technical Report

👍 4

08/13 08:00

Industrial recommender systems commonly use cascaded retrieval, ranking, and re-ranking pipelines. Although efficient, these pipelines fragment information and objectives across modules, rely on rigid rules, and have limited awareness of real-time intent, leaving session-level shifts among browsing,

Unlimited concurrency, pennies a minute

@TryLiveAvatar · 806 粉丝 · 973.2K 阅 · 504 赞 · 158 转

08/26 00:03

Why we removed LiveAvatar's concurrency limits and took prices down to $0.01 a minute. Starting today, LiveAvatar offers unlimited concurrency. You can run one avatar or ten thousand at the same time:

中文介绍 LiveAvatar取消并发限制，将价格降至每分钟0.01美元，实现无限并发，可同时运行一个或十个千个头像。

I gave my Grok Bot 3 X accounts. It got us 69.8M impressions. (full guide)

@chddaniel · 25.5K 粉丝 · 299.3K 阅 · 532 赞 · 26 转

08/24 23:00

69.8M impressions. 321K likes. 319K bookmarks. My brother David and I run these 3 X accounts as one content system. My Grok Bot is behind the ideas, the posts, the replies - everything. One post got

中文介绍 分享使用Grok Bot通过3个X账号获得6980万印象、32.1万点赞和31.9万收藏的完整指南。

What if everything goes right for AI? Learnings from the Aluminium trade.

@P_Bonnet · 5.7K 粉丝 · 262.7K 阅 · 519 赞 · 74 转

08/25 00:40

There are countless takes telling us why the AI buildout will result in a bubble that will collapse under its own weight. Some argue we are already in it, and that the subsequent meltdown will be

中文介绍 从铝贸易的角度探讨AI发展的可能性和潜在泡沫。

Designing Grok Bot with Grok Bot

@johnbai · 11.7K 粉丝 · 181.3K 阅 · 565 赞 · 29 转

08/25 06:40

At SpaceXAI, we’re building agents that can take an idea and begin making it real. For designers, that changes what’s worth exploring. Design has always been a process of giving form to an idea:

中文介绍 分享如何在SpaceXAI使用Grok Bot设计智能代理，将想法变为现实。

how i built an SEO/AEO blog engine

@harsehaj · 3.3K 粉丝 · 135.1K 阅 · 507 赞 · 35 转

08/24 08:55

i proposed and executed an engineering project end-to-end this summer as a growth engineering intern at @browserbase: blogEO, an engine that audits, rewrites, and generates SEO/AEO-tailored blogs on

中文介绍 分享在browserbase实习期间构建SEO/AEO博客引擎的工程经验。

How to Build a Company Brain That Gets Smarter Every Week

@VibeMarketer_ · 38.5K 粉丝 · 129.7K 阅 · 504 赞 · 50 转

08/25 21:31

I am going to show you how to turn every correction, decision, and useful workflow your team gives AI into one company brain that makes everyone better. Because right now, the smartest AI in your

中文介绍 展示如何将团队反馈转化为公司智能大脑，提高团队整体效率。

How to make money with Grok Bot

@EXM7777 · 132.1K 粉丝 · 126.3K 阅 · 504 赞 · 40 转

08/24 23:09

I'm going to show you how Grok Bot works, then hand you 10 workflows that drive real cash for your business... one specialized bot per workflow, running 24/7 on its own computer and i'll say it

中文介绍 展示Grok Bot如何运作，并提供10个工作流程，实现业务收益。

Odyssey Challenge Cinematic Guide

@imagine · 77.6K 粉丝 · 87.0K 阅 · 544 赞 · 73 转

08/27 02:08

It's been great to see everyone submit their Odyssey scene videos. Keep it up! The deadline for the Odyssey contest is August 31st. Meanwhile, here's a handy guide to help you get the most cinematic

中文介绍 提供Odyssey挑战赛电影场景视频的指导，帮助参赛者获得最佳效果。

Deep Dive: The Next Trillion-Dollar Futures Market

@chamath · 2.4M 粉丝 · 81.1K 阅 · 595 赞 · 52 转

08/25 00:49

“I actually believe a new asset class will be buying futures of compute. We just don’t have enough compute power right now.” - Larry Fink, CEO of BlackRock Recent news moved that prediction closer to

中文介绍 探讨下一个万亿美元的未来市场——计算力的期货市场。

What is harness engineering and why should I care?

@GoogleCloudTech · 1.3M 粉丝 · 35.3K 阅 · 526 赞 · 72 转

08/25 04:30

How do you ship a software product with 0 lines of manually-written code? A friend asked me this today, and I realized I didn’t have a simple answer. So I dug deeper. It turns out the answer is in how

中文介绍 探讨如何使用无代码技术进行软件产品交付，并介绍 Harness 工程的概念。

The Zero to Employable AI Engineering Roadmap

@_jaydeepkarale · 31.1K 粉丝 · 24.7K 阅 · 506 赞 · 86 转

08/24 21:46

A phase-by-phase path into AI engineering, with the specific courses, docs, and books worth your time at each stage, not just a list of skills to Google later Most AI engineering roadmaps are skill

中文介绍 分享从零到具备就业能力的AI工程师的成长路线图。

The inside story on why OpenAI agents hacked Hugging Face

MIT Tech Review AI · 08/27 03:00

The models responsible for last month’s agent hack of Hugging Face had been inadvertently trained to cheat and to communicate with each other, according to an OpenAI technical report released today. The hack, which a group of agents undertook to find solutions for a cybersecurity test that they were

中文介绍 OpenAI 技术报告显示，上个月 Hugging Face 的黑客攻击是由训练有误的模型执行的，这些模型被训练去作弊并相互通信。

Intelligent transcription with Gemini 3.5 Transcribe

DeepMind Blog · 08/27 01:01

Now you can get more intelligent speech-to-text transcription with Gemini 3.5 Transcribe.

中文介绍 DeepMind 推出 Gemini 3.5 Transcribe，提供更智能的语音转文字转录服务。

Lovable CTO: The Future of SaaS Is Apps That Agents Can Use

Latent Space · 08/27 00:16

Lovable is branching out from AI-powered web app creation and into MCP-powered ‘capabilities’. We talk to CTO Fabian Hedin.

中文介绍 Lovable 公司从 AI 驱动的网页应用创作扩展到 MCP 驱动的‘能力’，CTO Fabian Hedin 进行了讨论。

🔬“We have foundation models for language, not for physics” — Anima Anandkumar, Bren Professor of Computing

Latent Space · 08/26 23:15

Anima Anandkumar has spent two decades in AI, from classical math to deep learning and back. Now she's using it to model the physical world, from weather to fusion reactors.

中文介绍 Anima Anandkumar 教授在 AI 领域耕耘二十载，从经典数学到深度学习，现在用 AI 模拟物理世界。

Bringing ChatGPT for Teachers to more U.S. school districts

OpenAI News · 08/26 18:00

ChatGPT for Teachers is expanding to 55 U.S. school systems, bringing secure AI tools, training, and support to over 100,000 more educators and staff.

中文介绍 OpenAI 将 ChatGPT for Teachers 扩展到 55 个美国学区，为超过 10 万名教育工作者和员工提供安全 AI 工具、培训和支援。

Learning never stops: How AI makes learning continuous

OpenAI News · 08/26 18:00

OpenAI’s new report explores how students and educators use ChatGPT to make learning more continuous, with support that extends beyond the classroom.

中文介绍 OpenAI 新报告探讨了学生和教师如何使用 ChatGPT 使学习更加连续，支持延伸至课堂之外。

Raised on AI

MIT Tech Review AI · 08/26 17:00

When my oldest child was born, I immediately set up Gmail and Twitter accounts in her name. I broadly announced her birth online and proceeded to plaster her photo across all sorts of platforms. In short, I began creating her digital footprint long before she could stand on her own two feet. Fast-fo

中文介绍 作者在文章中讨论了在数字时代为新生儿创建数字足迹的现象。

AI models flub these intelligence tests. Can you fare any better?

MIT Tech Review AI · 08/26 17:00

Puzzles and games have been central to AI development since the very beginning. Just as we humans like to test our smarts with crosswords or logic puzzles, developers can test how far models have advanced with a gaming gauntlet. The term “machine learning” was popularized in a 1959 article by the IB

中文介绍 文章探讨了 AI 模型在智力测试中的表现，以及人类与机器学习之间的比较。

Bill Gates says we’ve passed AI’s danger thresholds. Now what?

MIT Tech Review AI · 08/26 15:01

It’s a glorious day in Kirkland, Washington, an affluent Seattle suburb on the eastern shore of Lake Washington. The temperature is in the mid-80s, and the sky is incapable of being any more blue. The view from the Gates Ventures conference room overlooks the Carillon Point Marina, where a flotilla

中文介绍 比尔·盖茨表示我们已经越过了 AI 的危险阈值，并讨论了接下来的应对措施。

The Hugging Face incident and the road ahead

OpenAI News · 08/26 08:00

OpenAI shares findings from the Hugging Face security incident and the steps we’re taking to strengthen AI model security, monitoring, and alignment.

中文介绍 OpenAI 分享了 Hugging Face 安全事件的调查结果，以及加强 AI 模型安全、监控和一致性的措施。

How loveholidays is making everyone a builder with Codex

OpenAI News · 08/26 08:00

Discover how loveholidays uses OpenAI Codex to make software development accessible across the business, helping teams turn ideas into products faster.

中文介绍 loveholidays 使用 OpenAI Codex 使软件开发在业务中变得可访问，帮助团队更快地将想法转化为产品。

Training and Finetuning Multi-Vector Embedding Models with Sentence Transformers

Hugging Face Blog · 08/26 08:00

中文介绍 使用 Sentence Transformers 训练和微调多向量嵌入模型。

OpenAI Jalapeño 🌶️, Perplexity Portable Computer  💻, Claude combines memory 🧠

TLDR AI · 08/26 08:00

中文介绍 OpenAI 推出 Jalapeño，Perplexity 可携式电脑，Claude 结合记忆功能。

Granite 4.2 LLMs: How They're Built

Hugging Face Blog · 08/25 23:14

中文介绍 介绍 Granite 4.2 LLMs 的构建方式。

Quantization-Aware Healing: a compressed, 4-bit model that outperforms its full-precision original

Hugging Face Blog · 08/25 19:39

中文介绍 介绍量化感知修复技术，这是一种压缩的 4 位模型，其性能优于原始全精度模型。