2026-07-31

TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM

👍 120

07/29 08:00

Vision-language-action (VLA) models commonly adopt an LLM-centric V to L to A pathway, where visual observations are projected into the representation space of a large language model before being decoded into robot actions. Although effective, this design incurs substantial computation and memory ov

中文介绍 提出TurboVLA，一种实时视觉-语言-动作模型，在RTX 4090上达到32 Hz推理速度且显存低于1 GB。通过重新设计VLA路径（将视觉观测投影到LLM空间再解码为动作），大幅减少计算和内存开销，实现高效实时机器人控制。

A New Role for Relevance: Guiding Corpus Interaction in Agentic Search

👍 87

07/27 08:00

Relevance is a query-dependent estimate of whether a document or excerpt contains useful evidence. Existing retrieval agents use relevance to select top-k content, but document relevance alone cannot localize, compose, or verify the evidence required by complex questions. Direct Corpus Interaction (

中文介绍 针对现有检索代理仅依赖文档相关性的局限，提出Direct Corpus Interaction（DCI）方法。DCI让代理主动与语料库交互，定位、组合和验证复杂问题所需证据，超越简单top-k检索，提升搜索效果。

CodeNib: A Multi-View Data System for Serving Repository Context to Coding Agents

👍 77

07/28 08:00

Coding agents repeatedly search, navigate, and retain context from evolving repositories, but disconnected indexes, language servers, and task-local histories force repeated discovery and obscure lifecycle costs. CodeNib builds reusable lexical, dense, and structural views per repository commit, map

中文介绍 提出CodeNib，一个为编码代理服务仓库上下文的多视图数据系统。它为每次仓库提交构建可复用的词法、稠密和结构视图，并映射到任务模板，使搜索、导航和上下文获取更高效，降低重复发现成本。

CoRT: Counterfactual Replay for Token-Level Rubric-Guided Policy Optimization

👍 74

07/28 08:00

Rubric-based reinforcement learning enriches language model training by evaluating model outputs against explicit criteria. Yet in GRPO-style pipelines, these structured judgments are reduced to a scalar response-level reward and converted into a response-level advantage, which is broadcast uniforml

中文介绍 针对GRPO-style pipeline中将结构化评分简化为标量奖励的缺陷，提出CoRT（反事实回放）。在token级别引入反事实优势信号，利用反事实推理提供更细粒度的反馈，提升语言模型训练的效率和效果。

ReDesign: Recovering Editable Design Structures from Images via Agentic Decomposition

👍 58

07/28 08:00

Recovering an editable design file from a raster image is a common and costly bottleneck in modern design workflows, yet remains challenging since editability depends on recovering multi-modal attributes, such as typography, vector geometry, colors, grouping, and layer ordering. We present ReDesign,

中文介绍 提出ReDesign，通过代理分解从光栅图像恢复可编辑设计结构。系统自动推断字体、向量几何、颜色、分组和图层顺序等多模态属性，使得编辑结构恢复成为可能，解决设计工作流中的关键瓶颈。

DecoEvo: Score-Decoupled Co-Evolution of Solver and Rubric-Generator Skills in Text Space

👍 54

07/28 08:00

Text-space optimization adapts large language models (LLMs) by editing external natural-language artifacts rather than model weights, so the optimized artifacts remain inspectable and the model can be treated as a black box. However, most existing text-space methods keep evaluation fixed. On open-en

中文介绍 提出DecoEvo，在文本空间中协同进化求解器和评分生成器。不修改模型权重，而是编辑自然语言工件（如提示、规则），同时优化求解策略和评分标准，在开放环境中实现可解释的黑盒优化。

CLBench-V: Evaluating Multimodal Context Learning from Grounding to Knowledge Acquisition

👍 40

07/28 08:00

Real-world tasks often require models to learn from task-specific context rather than relying only on pre-trained knowledge. While recent work has highlighted this capability as context learning, existing evaluations mainly focus on textual contexts. In many practical settings, however, the context

中文介绍 引入CLBench-V，评估多模态大模型从视觉上下文中学习的能力。覆盖从基础感知到知识获取的多个维度，填补现有评估仅关注文本上下文的空白，揭示模型在视觉情境学习中的表现和局限。

CAST: Game Solvers as Turn-Level Teachers for LLM Agents

👍 29

07/28 08:00

Training large language models (LLMs) to act in long-horizon games is a promising step toward generalist decision-making, yet reinforcement learning with verifiable rewards (RLVR) relies on sparse final rewards that reveal little about which decisions determine success. Denser process signals could

中文介绍 提出CAST，利用游戏求解器为LLM代理提供逐轮（turn-level）密集过程监督。相比稀疏的最终奖励，CAST在每个决策步骤给出细化反馈，有效提升长周期游戏中代理的强化学习训练效率。

Pass the Baton: Trajectory-Relayed On-Policy Distillation

👍 28

07/28 08:00

On-policy distillation (OPD) grounds token-level supervision in the student's own trajectory, yet suffers from prefix failure: once the student commits to a wrong reasoning direction, all subsequent generation builds on this deviation, producing misdirected continuations that elicit unreliable super

中文介绍 针对在线策略蒸馏（OPD）中前缀失败问题（学生一旦走偏，后续生成偏差），提出轨迹中继式在线策略蒸馏。通过中继正确推理轨迹，为后续token提供可靠监督，显著提升推理质量。

Keep It InMind: Benchmarking the Implicit-Association Blind Spot in Agent Memory

👍 28

07/27 08:00

Long-term memory systems store what a user says in an external store and retrieve it when a related query arrives. This interface rests on an assumption so natural that it is rarely stated: a memory that is needed will resemble the query that needs it. World knowledge breaks the assumption. A tree-n

中文介绍 揭示长期记忆系统的隐式关联盲点：现有系统假设需要的记忆与查询相似，但世界知识可能打破这一前提。提出Keep It InMind基准，评估代理在隐式关联场景中的记忆表现，挑战现有设计假设。

Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation Model

👍 26

07/27 08:00

Standard vision-language models (VLMs) suffer from Moravec's paradox: they excel at complex offline visual reasoning but struggle with simple streaming perception tasks and process them inefficiently. We present Mage-VL, an efficient codec-native streaming foundation model for real-time multimodal u

中文介绍 提出Mage-VL，一个编解码器原生的流式多模态基础模型。专门针对实时感知任务设计，克服标准VLM在流式场景下的低效问题，在保持视觉推理能力的同时实现高效流式处理。

SkillRise: Agentic Reinforcement Learning for Cross-Task Skill Evolution

👍 18

07/29 08:00

Large language model agents often encounter related yet distinct tasks that share reusable solution patterns. Yet standard agentic reinforcement learning treats tasks as independent episodes, while existing approaches to skill learning either focus on repeated attempts of one task or use pipelines w

中文介绍 针对代理强化学习将任务视为独立片段的问题，提出SkillRise。在相关但不同的任务间，通过agentic RL跨任务进化可复用技能，使得代理能积累和迁移解决方案模式，提升泛化能力。

MindForge: Teaching Small Language Models Whole-Life-Cycle Software Engineering via Source-Free Program Synthesis

👍 17

07/29 08:00

Coding agents have made substantial progress on software engineering tasks that modify existing codebases, including bug fixing and feature implementation. However, constructing a complete program from scratch remains a major challenge: even the frontier models evaluated on ProgramBench fully resolv

中文介绍 提出MindForge，通过源码无关的程序合成教授小语言模型全生命周期软件工程。不依赖现有代码，从零生成完整程序，解决从零编程的挑战，在ProgramBench上实现突破性解决率。

SpecFirst: Behavioral Specification Elicitation as a First-Class Step in Agent-Based Program Synthesis from Scratch

👍 15

07/29 08:00

LLM-based agents excel at software engineering tasks where an existing codebase provides context, but constructing a program from scratch remains fundamentally harder. Recent benchmarks such as ProgramBench quantify this gap: given only natural-language documentation and an execute-only binary as a

中文介绍 提出SpecFirst，将行为规范获取作为代理从零程序合成的一等步骤。先通过自然语言文档和二进制文件交互获取明确规范，再生成程序，有效提升从零编程的准确性和效率。

Visual prompt engineering for video models

👍 15

07/28 08:00

In the age of foundation models, a model is only as good as its prompt. For this reason, prompt engineering has become an essential technique for improving language model performance. Since video models are currently becoming foundation models for visual tasks (e.g., visual reasoning), we here ask w

中文介绍 研究视频模型的视觉提示工程。在视频基础模型时代，如何设计视觉提示（如边框、掩码）以引导模型输出。通过系统实验证明视觉提示能显著提升视频推理和生成任务性能。

StatePlay: State-Aware Game World Models for Mechanics-Consistent Generation

👍 13

07/29 08:00

Recent game world models can generate visually realistic and interactive environments conditioned on player actions. However, games are not defined by pixels alone; they are governed by explicit mechanics, namely state-dependent rules that control health reduction, skill activation, and game termina

中文介绍 提出StatePlay，状态感知游戏世界模型。在像素之外显式建模游戏状态依赖规则（如血量、技能激活），确保生成的交互环境遵循机制一致性，实现更逼真、可控的游戏世界生成。

Novel Claim or Déjà Vu? Rethinking "Contamination-Free'' Dynamic Evaluation for Multimodal Automated Fact-Checking

👍 13

07/26 08:00

Multimodal automated fact-checking (MAFC) verifies claims by retrieving and reasoning over external evidence. However, most existing static benchmarks risk contamination: they primarily consist of outdated claims verifiable using an LLM's internal knowledge without external evidence. This can inflat

中文介绍 针对多模态自动事实核查（MAFC）静态基准易受污染的问题，提出动态无污染评估方法。避免使用可通过LLM内部知识验证的旧声明，确保验证必须依赖外部证据，更真实反映MAFC能力。

PerceptionBench: Evaluating Atomic Visual Perception in Multimodal Large Language Models

👍 13

07/27 08:00

We introduce PerceptionBench, a benchmark specifically designed to evaluate the atomic visual perception capabilities of Multimodal Large Language Models (MLLMs). Existing benchmarks often fail to isolate perception: holistic evaluations conflate perceptual errors with failures in reasoning or domai

中文介绍 引入PerceptionBench，专门评估多模态大语言模型（MLLM）的原子视觉感知能力。通过隔离感知错误与推理失败，精确定位模型在颜色、形状、空间关系等基础视觉维度上的实际表现。

Can AI agents conduct open-ended AI research? Early evidence from two case studies

👍 12

07/29 08:00

Forecasts of explosive AI progress hinge on AI agents automating AI research. But evidence on whether agents can carry out open-ended AI research is thin. Current evaluations either test agents on narrow, verifiable tasks, which excludes open-ended research, or submit AI-generated papers to blind pe

中文介绍 通过两个案例研究，评估AI代理进行开放式AI研究的能力。现有测试仅限于窄验证任务，而人工评审又缺乏标准。初步证据表明代理能产生新颖想法但持续改进能力仍有限。

Parallel Decoding Distillation for Fast Image and Video Generation

👍 10

07/28 08:00

Generation in video diffusion or flow models is computationally expensive due to the slow and iterative sampling process. Current state-of-the-art (SOTA) acceleration methods heavily rely on variational score distillation (VSD) and adversarial losses to distill diffusion models into few-step generat

中文介绍 提出并行解码蒸馏（Parallel Decoding Distillation），用于加速图像和视频生成。结合变分分数蒸馏（VSD）和对抗损失，将扩散/流模型蒸馏为少步生成器，在保持质量的前提下显著降低采样耗时。

You already own the most valuable thing in shopping. You just can't reach it.

@fridayresearch_ · 1.9K 粉丝 · 1.6M 阅 · 597 赞 · 451 转

07/31 01:30

Your taste is scattered across forty companies that each own a piece and answer to advertisers. FRIDAY is building the version you own — a personal shopping intelligence that lives on your device and

中文介绍 FRIDAY 推出个人购物智能助手，收集用户散落在40家公司的品味数据，在设备端构建私有化购物智能，摆脱广告商依赖，强调数据所有权与隐私保护。

How to build an AI video studio in Claude Code:

@EXM7777 · 129.8K 粉丝 · 221.4K 阅 · 504 赞 · 36 转

07/28 22:13

I turned Claude Code into a working film studio and i'm going to hand you the complete system: the skills, the prompts, the loops, and the six-stage pipeline that ties them together each stage leans

中文介绍 使用者分享将 Claude Code 改造为 AI 视频工作室的完整方案，包含技能、提示词、循环和六阶段管线，实现全流程自动化视频制作。

What's gone wrong with AI & labor — a thought experiment

@random_walker · 128.4K 粉丝 · 134.0K 阅 · 552 赞 · 54 转

07/29 01:56

A thought experiment that I think helps explain much of what’s gone wrong with AI and labor: Imagine an alternate universe in which — for whatever reason — no one ever published source code online.

中文介绍 通过思想实验分析 AI 与劳动力问题：假设一个从未开源代码的平行宇宙，探讨开源与自动化如何改变就业格局，反思当前 AI 对劳动的影响。

The harness is all you need (mostly)

@github · 2.7M 粉丝 · 123.1K 阅 · 583 赞 · 69 转

07/29 04:28

A practical GitHub Copilot workflow for prototyping, planning, implementing, and reviewing software - without chasing every new AI tool. By @burkeholland If you’re feeling overwhelmed by AI right now,

中文介绍 GitHub 官方分享实用工作流：用 Copilot 完成软件原型的规划、实现和审查，无需追逐各种新 AI 工具，适合被工具疲劳困扰的开发者。

We rewrote our agent to run entirely in a Durable Object with Pi, Agents SDK and Code Mode

@Vercantez · 1.9K 粉丝 · 122.3K 阅 · 567 赞 · 44 转

07/29 00:19

We recently finished moving the camelAI agent off of virtual machines. The agent now runs inside a Cloudflare Durable Object, its filesystem lives in SQLite and R2, and it writes JavaScript instead of

中文介绍 camelAI 将代理从虚拟机迁移到 Cloudflare Durable Object，文件系统基于 SQLite 和 R2，改用 JavaScript 编写，实现云端无服务器架构的高效运行。

Here's exactly how to build your company brain (in 5 mins)

@DhravyaShah · 61.2K 粉丝 · 56.0K 阅 · 548 赞 · 36 转

07/28 14:16

Every company will have a company brain, whether they believe it or not. This is the bet that I'm making, and I'm constantly seeing very future forward startups come to us for their own setup. What is

中文介绍 博主分享 5 分钟内搭建公司大脑（企业知识库）的方法，预测每家公司都将拥有此类系统，已帮助多家前沿初创快速部署。

GPU Management: Why Idle GPUs Are the New Grounded Aircraft

Hugging Face Blog · 07/30 23:09

中文介绍 Hugging Face博客发文指出，闲置GPU如同停飞的飞机，造成资源浪费，呼吁加强GPU管理以提升利用率。

Gemini Robotics ER 2: powering robotics with video understanding, task orchestration, and multi-robot collaboration

DeepMind Blog · 07/30 23:00

Gemini Robotics ER 2 helps robots reason, collaborate, and solve real-world tasks. It represents a step change in video understanding, tool orchestration, and multi-robot collaboration for robotic applications.

中文介绍 DeepMind发布Gemini Robotics ER 2，通过视频理解、任务编排和多机器人协作，帮助机器人推理、协作并解决现实任务，是机器人应用的重要进步。

Ontologies Are So Back: Why AI Agents Are Reviving the Semantic Web

Latent Space · 07/30 19:17

AI engineers are rediscovering ontologies as a way to keep probabilistic agents inside deterministic boundaries.

中文介绍 Latent Space文章指出，AI工程师重新发现本体论，用于将概率性AI代理约束在确定性边界内，语义网因AI代理而复兴。

A fundamental flaw leaves LLMs strikingly vulnerable to attack

MIT Tech Review AI · 07/30 18:15

It is impossible to make large language models fully secure against hacks because of a fundamental flaw in how they work, a team of researchers argue in a paper presented at the International Conference on Machine Learning, a top AI conference, this month. The claim has huge implications for the saf

中文介绍 MIT Tech Review报道，研究人员在ICML会议上指出，由于工作原理的根本缺陷，大型语言模型无法完全防御黑客攻击，这一结论影响深远。

Advancing the price-performance frontier with GPT-5.6

OpenAI News · 07/30 18:00

Explore lower GPT‑5.6 pricing for Luna and Terra—and how OpenAI’s more efficient models help enterprises deploy AI workflows at scale.

中文介绍 OpenAI宣布GPT-5.6降低Luna和Terra定价，通过更高效率的模型帮助企业大规模部署AI工作流，推进性价比前沿。

OpenAI tops ARC-AGI-3 📈, GPT-5.6 efficiency ⚙️, AlphaFold team dissolution 🧬

TLDR AI · 07/30 08:00

中文介绍 TLDR AI简报：OpenAI在ARC-AGI-3基准测试中取得领先，GPT-5.6效率提升，AlphaFold团队解散。

[AINews] AI is eating Finance; AIE NYC now open

Latent Space · 07/30 07:32

a quiet day lets us cover how AI is permeating financial services as the next big vertical after coding.

中文介绍 Latent Space《AINews》报道，AI正在渗透金融服务领域，成为继编程之后的下一个重大垂直领域，同时AIE NYC现已开放。

We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control

DeepMind Blog · 07/30 00:02

中文介绍 DeepMind在Google Flow Music中推出Lyria 3.5，在音乐性、歌词、人声和创作控制方面取得进展。

How enabling two settings tripled our scores on the ARC-AGI-3 benchmark

OpenAI News · 07/29 23:00

How two API settings improved GPT-5.6 performance on ARC-AGI-3, boosting scores and efficiency by retaining reasoning and enabling compaction.

中文介绍 OpenAI宣布通过启用两个API设置，在ARC-AGI-3基准上使GPT-5.6性能提升三倍，通过保留推理和启用压缩提高得分和效率。

Accelerating scientific discovery with ChatGPT for Academic Researchers

OpenAI News · 07/29 18:00

OpenAI is giving 100,000 academic researchers free access to ChatGPT's most advanced AI models to accelerate scientific research, collaboration, and discovery.

中文介绍 OpenAI向10万名学术研究人员免费提供ChatGPT最先进AI模型访问权限，以加速科学研究、协作和发现。

The AI Hype Index: Unsexy AI

MIT Tech Review AI · 07/29 16:42

It feels bad enough when an open letter signed by leading economists warns that AI might steal your job. The fact it may soon be better than you at making dinner? Insult to injury. But that’s exactly what the company 1X promised when it showed off a pair of new, impressively dexterous (and, to some,

中文介绍 MIT Tech Review《AI Hype Index》聚焦不性感的AI，提及经济学家联名警告AI可能抢走工作，同时1X公司展示的新机器人具备惊人灵巧性，甚至能做晚餐。

not much happened today

Smol AI News · 07/29 13:44

**OpenAI's agent security incident expanded beyond Hugging Face, affecting four additional accounts and highlighting the need for stronger enterprise hardening measures like sandboxing and audit trails. The ongoing debate around "pacing the frontier" involves calls for coordinated slowdowns and gove

中文介绍 Smol AI News报道，OpenAI代理安全事件扩大，除Hugging Face外还有四个账户受影响，凸显沙箱和审计追踪等企业强化措施的必要性；同时关于放缓AI开发的辩论持续。

[AINews] Fearing RSI: OpenAI, Anthropic, GDM, Meta, Thinky cosign letter to "Pace" AI development, as HuggingFace details Machine-Speed Offensive Cyberattack

Latent Space · 07/29 08:46

The Big Pause is coming.

中文介绍 Latent Space报道，OpenAI、Anthropic、Google DeepMind、Meta等公司联名签署公开信，呼吁放缓AI开发步伐；同时Hugging Face披露机器速度网络攻击详情。

How GPT-5.6 fuses frontier intelligence with frontier efficiency

OpenAI News · 07/29 08:00

GPT-5.6 improves AI efficiency across models, inference, and agentic workflows, helping deliver more useful intelligence per dollar.

中文介绍 OpenAI介绍GPT-5.6如何融合前沿智能与前沿效率，在模型、推理和代理工作流中提升AI效率，实现每美元更多实用智能。

AI slowdown pact ⏸️, Personal superintelligence access 🌍, Grok Build Mode 🛠️

TLDR AI · 07/29 08:00

中文介绍 TLDR AI简报：多家公司签署AI放缓协议，个人超级智能访问权限开放，Grok推出构建模式。