2026-08-19

StreamOPD: A Post-Training Recipe with Spatio-Temporal Cue Gating for Streaming Video Understanding

👍 0

08/17 08:00

Streaming video understanding demands direct responses from the causally observed prefix of an unfolding video. Existing systems add inference-time memory, retrieval, and compression, yet a training-free sliding-window baseline already matches them. We therefore fix a memory-free recent-window proto

GRNEdit: Efficient General Video Editing from a New Binary-Evidence Perspective in Generative Refinement Networks

👍 13

08/17 08:00

Instruction-based general video editing seeks to unify diverse editing operations within a single, intuitive interface. Existing approaches often rely on resource-intensive conditioning, using either heavyweight branches or costly source concatenation. Is there any efficient way to model editing int

TRACE-Bench: Decomposing and Diagnosing Multi-Reference Image Generation

👍 9

08/17 08:00

Despite recent advances in unified multimodal models for multi-reference image generation, existing benchmarks remain organized around predefined task types (e.g., "subject composition"), which are ill-suited to this combinatorial setting and lead to fragmented coverage, uncontrolled complexity, and

Learn What's Left, Not What's Mastered: Saturation Aware Advantage Reweighting for Multi-Reward Policy Optimization

👍 44

08/17 08:00

Reinforcement learning (RL) with group-relative advantages has become the de facto standard for post-training language model reasoners. However, when optimizing multiple reward objectives, existing methods typically scalarize the reward vector with a fixed weighted sum before group-wise standardizat

AnyTalk: Speech Animation for Arbitrary Characters Leveraging a Video Generation Model

👍 5

08/17 08:00

We present AnyTalk, a novel method for generating 3D speech animations for arbitrary characters without requiring any animation data. While existing audio-driven 3D speech animation methods rely on character-specific training data or laborious rigging/re-meshing, AnyTalk circumvents these limitation

Prior Audit-Repair Context Shifts LLM Verifier Thresholds Toward Leniency

👍 3

08/17 08:00

Automated checking pipelines increasingly place one language model as the checker and another (or the same one) as the fixer. We ask whether that wiring changes what the checker reports. Measuring false alarms on human-verified-correct ProcessBench traces with the present task held byte-identical, w

HarnessEval-W: Agentifying the Evaluation of Visual Worlds

👍 111

08/17 08:00

A benchmark should deliver more than a scalar score: what makes an evaluation trustworthy is the reasoning that justifies the score. This is especially critical for world models, where judging a rollout requires understanding whether physics, causality, and world state evolve correctly. Humans spot

Improving the matrix multiplication exponent with modern optimization and AlphaEvolve

👍 10

08/17 08:00

The current best bounds on the matrix multiplication exponent ω are obtained through a refinement of the laser method called combination loss analysis (Duan et al., 2022; Williams et al., 2024; Alman et al., 2025). In this note, we address the optimization problem at the core of this approach and pr

Ventor-QTest: Threat-Model-Driven Verification of Vendor-Hosted LLM APIs

👍 12

08/17 08:00

As large language models become increasingly widespread, third-party providers that deploy open-weight models have become an important part of the ecosystem. Auditing the quality of their inference APIs is therefore an open problem. We formalize hosted model routing as a stochastic process and propo

ClawGym II: Exploring Black-Box RL on Agent Harness

👍 34

08/17 08:00

Agent harnesses have substantially improved performance on long-horizon tasks by coordinating agent interactions with the environment. However, reinforcement learning through complex harnesses remains largely unexplored, as scaling such training to long-horizon agent tasks introduces fundamental cha

R^3-Bench: LLMs Struggle with Resource-Rational Reasoning under Shared Budgets

👍 11

08/17 08:00

In cognitive science, resource rationality asks how an agent should allocate limited computation to maximize expected value. Most reasoning and agent benchmarks use independent per-task budgets; existing shared-budget studies do not calibrate suite performance against the same model's demonstrated s

GenRouter: Unified Workflow Routing for Agentic Image Generation

👍 19

08/17 08:00

The rapid evolution of text-to-image (T2I) generation models has effectively solved the foundational challenge of raw pixel synthesis, shifting the community's focus toward fulfilling increasingly intricate user requests. While recent agentic image generation workflows enhance static inference with

Beyond Visual CoT: Internalized Visual Thinking for Proactive Video Reasoning

👍 3

08/16 08:00

Multimodal large language models increasingly use visual chain-of-thought (Visual CoT) to reason about spatial, temporal, and embodied environments. By generating intermediate reasoning images, Visual CoT provides an intuitive mechanism for visual foresight but introduces substantial inference overh

Large Discovery Models: Empirically-grounded Model-Based Open-Ended Search

👍 49

08/16 08:00

Scientific discovery often involves optimising expensive-to-evaluate objectives over vast, structured, and open-ended hypothesis spaces, such as molecules, protein sequences, and computer programs. Generative models such as large language models (LLMs) provide expressive priors over such spaces, but

UI-Mate: Advancing Open-Weight Foundation GUI Agents with In-Context Demonstrations

👍 33

08/16 08:00

Foundation GUI agents can automate complex digital tasks, but deployment is hindered by scarce and biased training data, ambiguous prompts, and unreliable execution. Routine workflows rely on user-specific tools and tacit conventions, so unstated instructions can produce arbitrary variations across

ConceptFormer: Learning Adaptive Latent Concepts for Query-Document Alignment in Visual Document Retrieval

👍 4

08/16 08:00

Visual document retrieval is a critical component of multimodal retrieval-augmented generation, aiming to identify query-relevant pages from document collections where evidence is distributed across text, layout, charts, and visual structures. Recent efforts toward finer-grained supervision primaril

StateM: Reaching 95.3% Raw Accuracy, or a \$15 Frontier Run, on Terminal-Bench 2.1 via Harness Scaling

👍 110

08/15 08:00

Long-horizon agents can fail even when their underlying models can solve the constituent steps. They may lose track of mutable state, fail to reactivate lessons from earlier executions, skip known procedures, or stop prematurely. We bet on harness scaling to improve the execution system around an ag

MOSS-VL Technical Report

👍 38

08/15 08:00

We present MOSS-VL, an open vision-language model family that treats real-time interaction -- perceiving while it speaks -- as a first-class capability. It is co-designed across the stack: the language decoder attends to vision only through gated cross-attention, so the model can naturally see incom

VibeWorlding: Can Multimodal Agents Construct 3D Open Worlds End-to-End?

👍 51

08/15 08:00

Constructing an interactive 3D open world from a user query is important. However, existing methods are primarily evaluated on idealized, simple queries, making it difficult to systematically analyze and compare how multimodal agents understand user intent, use 3D tools, and reason over textual and

How Do Agents Fail on AutoResearch: End-to-End Diagnostic Evaluation on 100 Real-World Frontier Research Tasks

👍 22

08/14 08:00

AI has long assisted scientific research, but the rapid advance of LLMs and agentic scaffolds is reshaping the landscape; a single system can now carry whole-stage research from an initial hypothesis all the way to final published paper, which is a paradigm now referred to as AutoResearch. Existing

PACE-Bench: Benchmarking Physics Adaptation via Code Evolution in Dynamic Environments

👍 8

08/14 08:00

Self-evolving agents improve future behavior from interaction experience, yet existing evaluations typically optimize under fixed execution conditions and do not test recovery after those conditions change. To address this gap, we introduce PACE-Bench (Physics Adaptation via Code Evolution), a simul

Accuracy and Order Sensitivity Diverge Under Label-Free Strategies

👍 1

08/12 08:00

Multiple-choice benchmarks are widely used to evaluate large language models, but MCQ scores conflate knowledge with sensitivity to option order, which makes them unreliable measures of model knowledge. In this paper, we test whether preventing a model from seeing option labels while committing to a

VideoGAIA: A Benchmark for General AI Assistants on Agentic Video Understanding

👍 9

08/12 08:00

Video understanding is a fundamental task for evaluating the capabilities of multimodal large language models (MLLMs). However, existing leading models have already achieved approximately 90% accuracy on the Video-MME leaderboard, suggesting that conventional single-turn video understanding tasks ar

ENTLORE: A Graph-Grounded Benchmark for Latent Organizational Reasoning in Enterprise Question Answering

👍 4

08/11 08:00

Enterprise question answering is framed as retrieving internal documents and generating grounded answers. Routine enterprise records, however, are work by-products in which required organizational relations remain implicit across heterogeneous sources. Existing benchmarks provide realistic multi-sou

When Context Bites: Detecting RAG Poisoning via Document-Level Attention Collapse

👍 1

08/07 08:00

Retrieval-augmented generation (RAG) is indispensable for enhancing large language models. However, RAGs are increasingly susceptible to poisoning attacks, in which adversarial documents are injected to manipulate generator outputs. Previous methods rely on output-side signals such as perplexity and

Position: AI Agents in Scientific Teams Should Be Studied as Human-Agent Systems

👍 1

08/02 08:00

Large language model-based agents are increasingly deployed as collaborators in scientific discovery yet most current work focuses on the autonomous capabilities of "AI Scientists". We argue that this overlooks the social aspects of scientific teamwork, and that studying AI Scientists as human-agent

Valid Per-Field Selective Risk Control for Document Extraction: Three Failure Modes, a Validity Ladder, and When Conditioning Pays

👍 1

07/28 08:00

Per-field accept/review with selective risk at most alpha -- accept a field only if the error rate among accepted fields is controlled -- is the trust contract document-extraction systems need, and the natural procedure silently violates it on real documents. On 13,859 genuine claude-sonnet-5 fields

DumpsterCluster: From Dumpster Diving to Serving LLaMA-70B on $60 GPUs

👍 2

07/10 08:00

As AI datacenters retire functional GPUs, vast quantities of still capable accelerators enter secondary markets. This paper investigates whether these retired GPUs can find a productive afterlife to form a DumpsterCluster that can serve modern LLM inference, and under what conditions such repurposin

Plausible but Not Valid: A Psychometric Audit of LLMs as Synthetic Survey Respondents

👍 2

07/06 08:00

Large language models (LLMs) are increasingly used as synthetic survey respondents, but existing evaluations ask whether answers look plausible at the individual level. We argue the right question is psychometric: do LLMs preserve the joint distribution, latent structure, reliability, mediation path

HarmProfile: Characterizing Harmful Distributions in Frontier LLMs

👍 2

06/11 08:00

Frontier large language models (LLMs) safety evaluation has largely treated harmful generation as an attack outcome rather than as an object of analysis. Consequently, little is known about the harmful outputs produced during model misbehavior, partly because large-scale, high-quality collections of

How to build a $2M video production pipeline

@EXM7777 · 131.5K 粉丝 · 172.8K 阅 · 516 赞 · 38 转

08/16 22:12

i'm giving you 7 free skills and the exact pipeline directors are using to produce AI films with millions of dollars in budget in higgsfield... the whole system revealed step-by-step, with the skill

中文介绍 揭秘导演如何使用7个免费技能和详细流程制作耗资百万美元的AI电影。

Securing the Infrastructure of Intelligence

@JensenHuang · 1.0M 粉丝 · 110.9K 阅 · 1.9K 赞 · 234 转

08/17 20:40

Land, power and shell: The next critical resource for AI factories. AI factories are the defining infrastructure of the AI era—where compute transforms energy and data into intelligence that powers

中文介绍 探讨土地、能源和壳体成为AI工厂的关键资源，AI工厂是AI时代的核心基础设施。

Update to Pi App Studio Creation Pricing

@PiCoreTeam · 4.3M 粉丝 · 74.1K 阅 · 850 赞 · 222 转

08/18 06:19

Starting August 24, 2026, Pi App Studio will update how app creation and editing are priced. Until now, Pi has heavily subsidized the cost of creating and editing apps in App Studio for all creators.

中文介绍 Pi App Studio将于2026年8月24日更新应用制作和编辑的定价方式。

Grok Bot Masterclass

@_avichawla · 73.4K 粉丝 · 54.7K 阅 · 525 赞 · 48 转

08/18 18:14

Everything you need to understand, set up, and get real work out of Grok Bot. It covers what a Bot actually is, how the shared computer works underneath, how it compares to Hermes Agent, and how to

中文介绍 Grok Bot大师课，涵盖Bot的定义、共享计算机的工作原理、与Hermes Agent的比较以及如何使用。

Bring your spreadsheet data to life

@GoogleAI · 2.4M 粉丝 · 49.3K 阅 · 516 赞 · 53 转

08/18 03:39

Turn your spreadsheets into live, interactive mini-apps with Sheets canvas and Gemini 3.7 Flash — no coding required. How does it work? Let’s dive into it: Most AI UI tools produce prototypes that

中文介绍 Sheets canvas和Gemini 3.7 Flash将电子表格转换为互动迷你应用，无需编码。

Grok Bot for GTM

@kristaletz · 4.2K 粉丝 · 41.6K 阅 · 513 赞 · 24 转

08/17 05:34

This is how I use Grok Bot for Enterprise GTM at SpaceXAI. When we were initially dogfooding this product internally, Grok Bot rapidly gained adoption throughout the company, and I created an internal

中文介绍 分享如何在SpaceXAI使用Grok Bot进行企业级客户获取管理（GTM）。

Frontier Model Cost and Open-Weights Popularity is Driving Demand for Model Routing

Latent Space · 08/19 05:41

Glean CEO Arvind Jain explains why model routing helps control AI costs for organizations, and how human feedback loops at scale improve its routing systems.

中文介绍 Glean CEO Arvind Jain解释了模型路由如何帮助组织控制AI成本，以及大规模的人类反馈循环如何改善其路由系统。

Strengthening democratic oversight in national security

OpenAI News · 08/19 03:00

OpenAI launches an initiative to strengthen democratic oversight of AI in national security, supporting government institutions with tools, training, and expertise.

中文介绍 OpenAI推出一项加强国家安全领域AI民主监督的倡议，支持政府机构提供工具、培训和专业知识。

How Much Memory Does Your Agent Actually Need?

Hugging Face Blog · 08/19 02:09

中文介绍 探讨代理实际需要的内存量。

Partnering with CodeAI to prepare the first AI generation

OpenAI News · 08/18 19:00

OpenAI and CodeAI are partnering to help students build AI literacy, think critically about AI, and develop the skills to use and shape it responsibly.

中文介绍 OpenAI与CodeAI合作，帮助学生建立AI素养，批判性地思考AI，并发展负责任地使用和塑造AI的技能。

Pacing model development in an era of cyber-critical capabilities

OpenAI News · 08/18 19:00

OpenAI is strengthening monitoring, alignment, and security for frontier AI models. See how new safeguards are guiding the pace of model development.

中文介绍 OpenAI加强了对前沿AI模型的监控、对齐和安全，介绍了新的安全措施如何指导模型发展的步伐。

Introducing ChatGPT for Teens: Built for learning, backed by protections

OpenAI News · 08/18 19:00

ChatGPT for Teens helps teens learn, think critically, and use AI with confidence, with stronger built-in protections, healthy-use features, and additional controls for parents.

中文介绍 ChatGPT for Teens帮助青少年学习、批判性思考和使用AI，拥有更强的内置保护、健康使用功能和家长控制。

We still don’t know how people are really using AI

MIT Tech Review AI · 08/18 18:06

AI companies like Anthropic and OpenAI regularly publish reports on how people are using products like Claude and ChatGPT, but they only release the data they want us to see, AI researchers say. “There is no independent source to corroborate it,” says Anka Reuel, a computer science PhD candidate at

中文介绍 AI研究人员表示，尽管AI公司如Anthropic和OpenAI定期发布关于人们如何使用Claude和ChatGPT等产品的报告，但他们只发布我们想看到的数据，没有独立的来源来证实。

AI’s recursive self-improvement might not come so quickly after all

MIT Tech Review AI · 08/18 17:00

The AI industry’s boldest promise right now is that AI will soon improve itself, with almost no need for human oversight. LLMs can already write code, generate synthetic data for training, and optimize the computer chips they run on. Forecasts of explosive AI progress predict that what researchers c

中文介绍 AI行业目前最大胆的承诺是AI将很快自我改进，几乎不需要人类监督。

Asana cleared 5 years of engineering work in 2 weeks with Codex

OpenAI News · 08/18 15:00

Asana used OpenAI Codex to replace an outdated testing system in two weeks, completing work expected to take five years for about $12K.

中文介绍 Asana使用OpenAI Codex在两周内替换了过时的测试系统，完成了原本需要五年时间的工作，花费约12,000美元。

How NVIDIA scales expertise with ChatGPT Work

OpenAI News · 08/18 08:00

NVIDIA teams use ChatGPT Work to reduce manual tasks, connect fast-moving signals, and scale successful workflows globally.

中文介绍 NVIDIA团队使用ChatGPT Work减少手动任务，连接快速移动的信号，并在全球范围内扩展成功的流程。

Multi-Vector (Late Interaction) Embedding Models with Sentence Transformers

Hugging Face Blog · 08/18 08:00

中文介绍 介绍使用Sentence Transformers的多向量（后期交互）嵌入模型。

Cursor Origin 👨‍💻, Anthropic $65B revenue 💰, deadline dividend scaling 📈

TLDR AI · 08/18 08:00

中文介绍 Cursor Origin，Anthropic 650亿美元收入，最后期限规模扩大。

[AINews] Stripe buys OpenRouter for $7B

Latent Space · 08/18 07:13

No GPUs, no Agents, just really, really, really good infra and distribution.

中文介绍 Stripe以70亿美元收购OpenRouter，无GPU、无代理，只有非常好的基础设施和分销。

Same Cluster, 33 Points More Utilization: What Changed Was the Order

Hugging Face Blog · 08/18 03:46

中文介绍 同一集群，33点更多利用率：变化的是顺序。

What Flock’s defenders are missing

MIT Tech Review AI · 08/18 03:16

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Flock, the police-tech giant known for its network of some 120,000 automatic license plate readers around the US, announced some changes to its platform last Thur

中文介绍 Flock的捍卫者所忽视的问题。