2026-07-03

EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments

👍 39

07/02 08:00

Autonomous agents are increasingly expected to improve executable policies through feedback, yet existing evaluations often collapse this process into a final score or confound it with open-ended software-engineering progress. We introduce Autonomous Policy Evolution, a controlled evaluation setting

中文介绍 提出 Autonomous Policy Evolution 可控评估框架，用于衡量智能体在交互环境中通过反馈改进可执行策略的能力，避免将过程压缩为单一分数或与软件工程进展混为一谈。

AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents

👍 38

07/02 08:00

Memory for a long-horizon LLM agent is a contract about what each future decision is allowed to see. The simplest contract appends past observations, tool calls, and reflections to every prompt, which makes prior context easy to access but also turns it into a jumbled mixture in which the effect of

中文介绍 提出 AgenticSTS 有界内存测试床，用于评估长程 LLM 智能体的记忆设计。指出简单拼接所有上下文导致混乱，并为未来记忆契约研究提供基准。

PerceptionRubrics: Calibrating Multimodal Evaluation to Human Perception

👍 36

06/26 08:00

We introduce PerceptionRubrics, a rubric-based evaluation framework that addresses the gap between saturated benchmark scores and real-world brittleness. Shifting evaluation from holistic semantic matching to rigorous atomic auditing, PerceptionRubrics pairs 1,038 information-dense images with over

中文介绍 提出 PerceptionRubrics 量规评估框架，将多模态评估从整体语义匹配转向原子化审计，配以 1,038 张信息密集图像和超过 10,000 个细粒度问题，校准模型表现与人类感知。

Morphing into Hybrid Attention Models

👍 29

06/29 08:00

Hybrid attention models improve long-context efficiency by retaining only a subset of full-attention layers and replacing the remaining layers with linear attention. However, the effectiveness of Transformer-to-hybrid conversion critically depends on which layers preserve full attention. Existing hy

中文介绍 提出 Morphing 方法，自动选择 Transformer 中哪些层保留全注意力、哪些替换为线性注意力，以优化长上下文效率。相比随机或固定选择策略显著提升性能。

TurboServe: Serving Streaming Video Generation Efficiently and Economically

👍 27

06/17 08:00

Streaming video generation is emerging as a new serving workload in which users interact with long-lived sessions that generate video progressively, chunk by chunk. Unlike offline video generation or typical LLM serving, streaming video generation must preserve session state across active and idle p

中文介绍 提出 TurboServe 系统，高效服务流式视频生成，通过保存跨空闲和活跃期的会话状态，减少计算冗余，显著降低成本并保持低延迟。

Multimodal Continuous Reasoning via Asymmetric Mutual Variational Learning

👍 22

07/01 08:00

Multimodal Large Language Models (MLLMs) are often constrained by a language-space bottleneck, forcing complex visual reasoning into discrete tokens which can lose perceptual nuance. A promising alternative is continuous latent reasoning, where the goal is to discover implicit reasoning pathways tha

中文介绍 提出非对称互变分学习框架，实现多模态连续潜在推理。通过非对称编码器-解码器结构，在潜在空间中联合建模视觉和语言，避免离散 token 损失，提升细粒度推理能力。

ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving

👍 22

07/01 08:00

In prefill-decode (PD) disaggregated LLM serving, each request is assigned to a decode worker after prefill. Existing decode routers balance only load; for mixture-of-experts (MoE) models this is incomplete: equally loaded workers can differ in latency, since each decode step loads the weights of ev

中文介绍 提出 ELDR 专家局部感知路由算法，用于预填充-解码分离的 MoE 服务。在路由时考虑专家权重的局部性，减少跨 worker 的专家加载开销，相比纯负载均衡降低解码延迟。

Seed2.0 Model Card: Towards Intelligence Frontier for Real-World Complexity

👍 22

06/30 08:00

We present Seed2.0, a model series that takes a meaningful step toward solving complex, real-world tasks. Our approach begins with identifying users' genuine needs and constructing a reliable, forward-looking evaluation system by selecting and abstracting benchmarks grounded in these needs and in re

中文介绍 介绍 Seed2.0 模型系列，针对复杂真实世界任务设计，构建基于用户需求的评估体系，在推理、工具使用和多模态能力上达到前沿水平。

MemSyco-Bench: Benchmarking Sycophancy in Agent Memory

👍 21

07/01 08:00

Memory has emerged as a cornerstone of modern LLM-based agents, supporting their evolution from single-turn assistants to long-term collaborators. However, memory is not always beneficial: retrieved memories often induce a critical issue of sycophancy, causing agents to over-align with the user at t

中文介绍 提出 MemSyco-Bench 基准，系统评估 LLM 智能体在记忆影响下的谄媚行为。设计多种记忆场景和评估指标，揭示记忆加剧过度迎合用户的倾向。

AgenticDataBench: A Comprehensive Benchmark for Data Agents

👍 19

07/02 08:00

Data science aims to derive actionable insights from heterogeneous raw data, unlocking the value of the massive amounts of data generated in modern society. Automating this process is essential to reducing labor-intensive efforts for data scientists and enabling scalable data-driven applications. Re

中文介绍 提出 AgenticDataBench 全面基准，评估 LLM 数据智能体在数据科学任务上的能力，覆盖数据探索、清洗、分析和可视化，提供标准化评估。

ASPIRE: Agentic /Skills Discovery for Robotics

👍 16

06/30 08:00

Traditional robot programming is challenging: it requires orchestrating multimodal perception, managing physical contact dynamics, and handling diverse configurations and execution failures. We introduce ASPIRE (Agentic Skill Programming through Iterative Robot Exploration), a continual learning sys

中文介绍 提出 ASPIRE 自主技能探索系统，使机器人通过持续探索自主发现和复用操作技能，减少手工编程，在多种任务上提升泛化能力。

WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory

👍 15

07/02 08:00

We present WorldDirector, a highly controllable video world model framework designed for persistent dynamic object memory and unrestricted viewpoint exploration. Unlike existing world models that entangle physical dynamics with pixel rendering and rely on continuous visual observation to sustain mot

中文介绍 提出 WorldDirector 可控视频世界模型框架，引入持久动态对象记忆，解耦物理模拟与像素渲染，支持自由视角探索和对象状态持久化，提升交互一致性。

Breaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal Reasoning

👍 13

06/30 08:00

Recent multimodal large language models have shown great promise in clinical image reasoning, but existing post-training pipelines remain predominantly outcome-centric, relying on final answer correctness or sequence-level preferences. This suffers from sparse credit assignment, making it difficult

中文介绍 提出 Step-Aware RL 方法，对医疗多模态推理的每个步骤进行细粒度奖励和优化，打破错误级联。在临床影像问答上显著优于仅关注最终答案的强化学习。

AutoTrainess: Teaching Language Models to Improve Language Models Autonomously

👍 13

06/30 08:00

Training language models (LMs) remains a highly human-intensive process, even as frontier language model agents become increasingly capable at software engineering and other long-horizon tasks. A central challenge is that autonomous post-training is not just a coding problem: it requires the agent t

中文介绍 提出 AutoTrainess 框架，让语言模型自主管理后训练流程，包括数据生成、超参数调优和验证，大幅减少人工干预，实现自我改进循环。

Perceive-to-Reason: Decoupling Perception and Reasoning for Fine-Grained Visual Reasoning

👍 12

07/01 08:00

Fine-grained visual reasoning remains challenging for vision-language models, especially when small but critical visual cues are buried in high-resolution images. Existing approaches rely on repeated cropping or test-time visual search to introduce local evidence, but they typically do not explicitl

中文介绍 提出 Perceive-to-Reason 方法，先通过感知模块定位关键视觉区域，再单独进行推理，解耦感知与推理过程，提升细粒度视觉推理准确性。

SkillCoach: Self-Evolving Rubrics for Evaluating and Enhancing Agentic Skill-Use

👍 11

07/02 08:00

Skills are becoming a reusable operational layer for LLM agents, encoding SOPs, domain rules, tool workflows, scripts, and validation routines. In realistic skill repositories, overlapping skills make reliable skill-use difficult. Final verifier success is too coarse for both evaluation and training

中文介绍 提出 SkillCoach 自我演化量规，用于评估和提升 LLM 智能体对技能的正确选择与使用。通过细粒度反馈自动修正技能调用，超越最终结果验证。

Logit-Contribution Scoring Identifies Non-Literal Retrieval Heads

👍 9

07/01 08:00

In long-context use, large language models frequently synthesize answers from the meaning of a relevant context span rather than literally copy-pasting them. Identifying which attention heads perform this synthesis matters for interpreting long-context model behavior. Yet existing detectors miss the

中文介绍 提出 Logit-Contribution 评分方法，识别 LLM 中执行非字面检索（语义合成）的注意力头，为长上下文解释提供新工具，揭示此前检测器遗漏的机制。

The State-Prediction Separation Hypothesis

👍 8

07/01 08:00

Transformers use the same forward computation stream to both predict the next token and store useful state for future token predictions. We formulate the state-prediction separation hypothesis: disentangling the two roles yields better language modeling performance. We design a Transformer variant t

中文介绍 提出状态-预测分离假设：将 Transformer 的下一词预测与隐状态存储功能解耦。设计分离架构变体，在语言建模上取得更好性能，验证了该假设。

InstanceControl: Controllable Complex Image Generation without Instance Labeling

👍 7

06/30 08:00

Controllable image generation methods, such as ControlNet, have demonstrated a remarkable capacity to introduce visual conditions(e.g., depth maps) to guide image generation. However, these methods often struggle with complex multi-instance scenes, frequently leading to attribute confusion among ins

中文介绍 提出 InstanceControl 方法，无需实例级标注即可实现复杂多实例图像的可控生成。通过解耦实例级条件，缓解属性混淆问题，在多个多实例场景中提升一致性。

When Search Agents Should Ask: DiscoBench for Clarification-Aware Deep Search

👍 7

06/26 08:00

Search agents powered by large language models (LLMs) are increasingly used to solve complex information-seeking tasks, requiring multi-step retrieval and reasoning to fulfill user goals. However, existing benchmarks often assume that user queries are complete and explicit, overlooking the fact that

中文介绍 提出 DiscoBench 基准，评估搜索智能体主动提出澄清问题的能力。包含需要澄清的查询集合，衡量何时提问能提升搜索效率和最终答案质量。

PorTAL: Portable Task Adapters for LLMs

@RampLabs · 13.4K 粉丝 · 335.4K 阅 · 507 赞 · 45 转

07/02 02:09

Researcher: Ben Geist Abstract Parameter-efficient fine-tuning (e.g. LoRA) adapts a frozen LLM to a task, but the resulting adapter is locked to one base model. When a new model is released, the

中文介绍 介绍PorTAL方法，解决LoRA微调适配器只能绑定单一基座模型的问题。通过可迁移任务适配器，允许适配器在不同LLM间复用，提升参数高效微调的灵活性。

How To Master Fable (Fundamentals Guide)

@milesdeutscher · 671.1K 粉丝 · 243.8K 阅 · 500 赞 · 73 转

07/02 02:13

TL;DR: Everything you need to do to get maximum value from Fable. I guarantee that after you're done reading, you'll have all the necessary tools to quite literally 10x your AI productivity with

中文介绍 分享Fable工具的全面使用指南，从基础到进阶，帮助用户最大化AI生产力。涵盖核心功能、操作技巧，目标是让AI工作效率提升10倍。

Career advice in the age of AI

@philhchen · 9.1K 粉丝 · 179.1K 阅 · 516 赞 · 34 转

07/03 05:25

AI models get better at anything you can write a loss function for, and school is mostly loss functions: well-defined problems graded against known answers. Therefore, the valuable work of the next

中文介绍 AI时代职业建议：学校训练的多为可量化的“损失函数”问题，而未来有价值的工作是那些无法用损失函数定义的创造性、模糊性任务。

Toward Unrestricted Intelligence: Venice Series A

@ErikVoorhees · 908.5K 粉丝 · 124.8K 阅 · 529 赞 · 103 转

07/01 22:28

“If others possess your thoughts and constrain your words, then you exist at their permission, and you are not free.” —Anonymous Venice launched just over two years ago to create a private and

中文介绍 Venice项目宣布A轮融资，致力于打造私密、不受限制的AI平台。强调思想自由和语言自由，为用户提供无审查的AI对话环境。

46 thoughts on the near future

@bayeslord · 41.6K 粉丝 · 86.2K 阅 · 511 赞 · 43 转

07/01 04:37

This list is based on a thread I posted on June 4th. A few edits and additions here and there. Several people asked me to make the thread easier to read, so here it is. Intelligence I think people are

中文介绍 整理46条关于AI与近未来的深度思考，涵盖智能发展、社会变革等议题。源自作者之前推文，经编辑后形成更易读的列表。

THE MOST VALUABLE THING YOU CAN DO WITH FABLE 5 IN THE NEXT 24 HOURS

@AlexFinn · 459.5K 粉丝 · 72.5K 阅 · 565 赞 · 34 转

07/03 08:20

If the first thing you did with Fable 5 was vibe code, you're using it wrong. Fable 5 isn't a vibe coding tool. It's an operating systems tool. Let me explain. Fable 5 is the first model I've ever

中文介绍 Fable 5的首要用途不是vibe coding，而是作为操作系统工具。建议用户将模型用于更深层的系统级工作，而非简单开发。

What I learned helping set up 200 company brains

@jacob_posel · 28.7K 粉丝 · 54.5K 阅 · 501 赞 · 30 转

06/30 23:59

I have been directly or indirectly involved in helping set up over 200 company brains and context layers. Here are the three most important lessons. 1. The Biggest Predictor Of Success Is An Internal

中文介绍 总结为200家公司搭建知识库/上下文层的经验。三大教训：内部上下文层是成功最大预测因素，另两个关键教训涉及组织协作。

What Are Agents Paying For?

@base · 1.4M 粉丝 · 51.0K 阅 · 501 赞 · 112 转

07/01 14:43

Written by: @Must_be_Ash You've heard the slogans: "the agentic economy is here" and "agents are becoming the internet's newest paying customers." They sound big, but they skip the obvious question:

中文介绍 分析AI Agent经济中实际支付的对象。指出常见口号忽略的关键问题，探讨Agent在工具、API、计算等方面的消费模式。

Fable is back, here's how I use it in Cursor

@ericzakariasson · 76.4K 粉丝 · 37.9K 阅 · 567 赞 · 24 转

07/02 19:10

Fable is back in Cursor, and here's a pattern I've been exploring and some other ways I've been getting the most out of the model. Fable as orchestrator, Composer as workers It's easy to put

中文介绍 展示Fable模型在Cursor编辑器中的使用模式：将Fable作为编排器，Composer作为工作线程。利用两者协作提升开发效率。

Your One-Page PyTorch Training Pipeline Cheat Sheet.

@0xkozue · 690 粉丝 · 24.9K 阅 · 501 赞 · 46 转

07/02 17:03

The entire engine of Deep Learning works by making tiny, continuous adjustments to a model's weights. This is my PyTorch training pipeline cheat sheet. If you spot any mistakes or have suggestions for

中文介绍 提供一页PyTorch训练流程速查表，涵盖权重调整、前向传播、损失计算等核心步骤。适合深度学习开发者快速参考。

Building Infra for the Agent-Native Internet: June Recap, July Vision

@OptimaiNetwork · 99.8K 粉丝 · 7.8K 阅 · 540 赞 · 351 转

07/01 23:05

The AI industry is entering a new phase. The conversation is no longer centered on models alone. Increasingly, it is about infrastructure: how intelligent systems access knowledge, coordinate with one

中文介绍 讨论AI行业从模型转向基础设施的趋势。回顾六月进展，展望七月愿景，重点聚焦智能系统如何访问知识、协调协作。

Google DeepMind and A24 announce first-of-its-kind research partnership

DeepMind Blog · 07/03 22:25

中文介绍 Google DeepMind 与电影公司 A24 宣布建立首个此类研究合作伙伴关系。

AIEWF Daily Dispatch: The great loops debate and the state of AI engineering

Latent Space · 07/03 13:11

The AI Engineer World’s Fair ended with a debate about loops, a report on the state of AI engineering, and closing keynotes focused on what to build next.

中文介绍 AI工程师世界博览会以关于循环的辩论、AI工程现状报告和关于下一步构建内容的主题演讲结束。

Vercel's Andrew Qu on why agents are a new kind of software

Latent Space · 07/03 08:08

The Vercel Chief of Software explains how its agent framework, eve, was created — and why skills, sandboxes and agent-readable websites now matter.

中文介绍 Vercel 软件主管 Andrew Qu 解释其智能体框架 eve 的创建过程，以及技能、沙箱和智能体可读网站为何重要。

Meta Watermelon 🍉, Anthropic Samsung chips 🤝, autoresearch in practice 📈

TLDR AI · 07/03 08:00

中文介绍 TLDR AI 报道：Meta 的 Watermelon 项目、Anthropic 与三星芯片合作、自动研究实践。

The website of the future may assemble itself for every visitor

Latent Space · 07/03 05:25

Adobe is experimenting with “agentic sites” that generate pages around an individual user’s intent. At AIEWF, we talked to Carlos Sanchez about the Web's future.

中文介绍 Adobe 正在实验“智能体网站”，可为每个用户意图生成页面。Carlos Sanchez 在 AIEWF 上讨论网络未来。

Achieving operational excellence with AI

MIT Tech Review AI · 07/02 23:37

Frameworks like Lean Six Sigma and business process management (BPM) first gained traction because they promised clarity in the chaos—a structured way to bring order to messy, sprawling operations. Lean Six Sigma emphasized statistical rigor and quality control; BPM created end-to-end maps of how wo

中文介绍 Lean Six Sigma 和业务流程管理（BPM）等框架曾为混乱运营带来秩序，AI 正推动运营卓越。

Skill engineering and the case against one-shot AI design

Latent Space · 07/02 22:36

Paul Bakaus talks to us about Impeccable, human judgment in a 'loopmaxxing' era, and why agents still need people to steer them.

中文介绍 Paul Bakaus 讨论 Impeccable 以及人类判断在“循环最大化”时代的重要性，指出智能体仍需人类引导。

Teaching AI to run with the turbines

MIT Tech Review AI · 07/02 20:51

Artificial intelligence may have captured the public imagination through chatbots and image generators, but some of its most consequential use cases are unfolding far from consumer-facing tools. In industries where physical infrastructure, operational continuity, and safety are paramount, AI is beco

中文介绍 AI 的最重要应用远非聊天机器人，而是在物理基础设施、运营连续性和安全领域，如涡轮机运维。

[AINews] not much happened today

Latent Space · 07/02 15:10

another quiet day.

中文介绍 当天新闻平淡。

AIEWF Daily Dispatch: Autoresearch and the tension between AI and human agency

Latent Space · 07/02 14:13

The software factory vision met resistance today from speakers defending human understanding and control.

中文介绍 AIEWF 每日快报：软件工厂愿景遭遇反对，演讲者维护人类理解和控制权。

not much happened today

Smol AI News · 07/02 13:44

**Fullstack Code Arena** extends coding agent evaluation to include **databases, API keys, deployments, and structured tool use**, marking a shift to end-to-end app shipping. **LangChain** released **LangSmith** with unified tracing and **OpenWiki** for auto-generated docs, while **LlamaIndex** demo

中文介绍 Fullstack Code Arena 扩展编码智能体评估至数据库、API、部署；LangChain 发布 LangSmith 和 OpenWiki。

Gemini Flash upgrade ⚡️, Meta AI cloud 🌐, ZCode 👨‍💻

TLDR AI · 07/02 08:00

中文介绍 TLDR AI 报道：Gemini Flash 升级、Meta AI 云服务、ZCode 项目。

Autoresearch: The feedback loop behind self-improving agents

Latent Space · 07/02 07:52

Introspection co-founder Roland Gavrilescu explains autoresearch, agent “recipes,” self-improving loops, and why humans remain central to the software factory.

中文介绍 Introspection 联合创始人 Roland Gavrilescu 解释自动研究、智能体“配方”、自我改进循环，强调人类在软件工厂中的核心地位。

How Cursor deploys AI inside the enterprise

Latent Space · 07/02 03:03

Cursor's Pauline Brunet explains how her team of Forward Deployed Engineers help organizations implement agents — essentially setting up software factories.

中文介绍 Cursor 的 Pauline Brunet 解释其现场部署工程师团队如何帮助企业实施智能体，建立软件工厂。

🔬 The Coolest Diffusion Research Isn't in LLMs — Evan Feinberg & Sergey Edunov, Genesis Molecular AI

Latent Space · 07/01 22:42

Why the Llama lead left Meta for drug discovery, PEARL's zero-shot OpenBind win, and what becomes possible when co-folding finally crosses the accuracy threshold.

中文介绍 Llama 负责人离开 Meta 投身药物发现，PEARL 零样本赢得 OpenBind，共折叠达到精度阈值带来新可能。