2026-07-05

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

👍 44

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，一个有界记忆测试床，用于评估长周期LLM agent的记忆契约。通过限制记忆容量并分析不同记忆策略对决策的影响，揭示了简单拼接历史信息导致上下文混乱的问题，为长周期agent的memory设计提供评估基准。

EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments

👍 41

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EvoPolicyGym，一个在交互环境中评估自主策略演化的框架。区别于传统最终得分或开放式软件工程进展，该工作将策略演化过程从其他因素中分离，提供可控评估，用于衡量agent通过反馈改进可执行策略的能力。

PerceptionRubrics: Calibrating Multimodal Evaluation to Human Perception

👍 38

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，一个基于评分细则的多模态评估框架，将评价从整体语义匹配转向原子化审计。配对1,038张信息密集型图像与超过（略），目的是校准评估以符合人类感知，弥补基准饱和与实际表现间的差距。

Morphing into Hybrid Attention Models

👍 37

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 研究Transformer到混合注意力模型的转换，提出一种确定哪些层应保留全注意力的方法。解决现有混合模型在保持长上下文效率时性能下降的问题，通过层选择优化实现高效注意力分配。

AgenticDataBench: A Comprehensive Benchmark for Data Agents

👍 25

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，一个全面的数据智能体基准，用于评估LLM代理在数据科学自动化任务中的表现。涵盖数据清洗、查询、分析等环节，提供标准化评估，推动数据科学自动化发展。

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

👍 24

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，一种感知专家局部性的解码路由策略，用于prefill-decode分离的MoE模型服务。通过考虑各工作节点上专家权重的缓存情况，在均衡负载的同时最小化解码延迟，提升MoE服务效率。

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

👍 23

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 发布Seed2.0模型系列，向解决复杂现实世界任务迈出重要一步。基于用户真实需求构建可靠评估体系，模型在多领域展现前沿能力。该模型卡详细描述了模型架构、训练数据及评估结果。

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实现多模态连续推理，突破MLLM的语言空间瓶颈。通过连续潜在推理路径捕捉视觉感知细节，避免离散化信息损失。非对称变分框架促进多模态对齐，提升复杂视觉推理性能。

MemSyco-Bench: Benchmarking Sycophancy in Agent Memory

👍 22

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，用于评测LLM agent记忆中存在的奉承问题。记忆检索可能导致agent过度迎合用户偏好，该基准设计了特定场景评估这一偏差，为构建更可靠的长时记忆agent提供测试工具。

WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory

👍 20

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，一个高度可控的视频世界模型框架，具备持久动态对象记忆和自由视角探索能力。通过分离物理动态与像素渲染，支持对运动对象的长期跟踪和场景操作，实现更可信的世界模拟。

ASPIRE: Agentic /Skills Discovery for Robotics

👍 18

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（基于迭代机器人探索的自主技能编程），一个持续学习系统，使机器人通过自主探索发现和习得技能。无需人工编程，就能处理多模态感知和复杂动力学，实现机器人技能的自动化获取和组合。

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

👍 17

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 提出步骤感知强化学习方法用于医学多模态推理，解决结果导向训练中的稀疏信用分配问题。通过逐步奖励和中间步骤监督，打断失败级联，提升临床影像推理的准确性和鲁棒性。在医学数据集上取得显著改进。

Logit-Contribution Scoring Identifies Non-Literal Retrieval Heads

👍 14

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贡献评分方法，识别LLM中执行非字面检索的注意力头。在长上下文场景中，模型常基于语义合成答案，该方法通过分析注意力头的logit贡献，准确定位负责语义检索的头，提升模型可解释性。

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

👍 14

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，利用自演进的评分细则来评估和增强LLM agent的技能使用能力。针对技能库重叠问题，通过构建细粒度评价指标和自动修正反馈，提升agent在复杂任务中选择和组合技能的可靠性。

AutoMem: Automated Learning of Memory as a Cognitive Skill

👍 12

07/01 08:00

Memory expertise is a learned skill: knowing what to encode, when to retrieve, and how to organize knowledge--a capacity known in cognitive science as metamemory. We bring this perspective to LLMs by treating memory management as a trainable skill. We promote file-system operations to first-class me

中文介绍 提出AutoMem，将记忆管理训练为LLM的认知技能。受元记忆启发，将文件系统操作提升为第一类元技能，使模型学会自主决定何时编码、检索和组织知识，实现对长周期对话中记忆的高效利用。

Graph-Native Reinforcement Learning Enables Traceable Scientific Hypothesis Generation through Conceptual Recombination

👍 9

07/01 08:00

Accelerating materials discovery requires AI systems that can generate scientifically valid hypotheses through multi-step, domain-grounded reasoning. Standard large language models often produce fluent but weakly traceable responses to open-ended materials design problems, making it difficult to det

中文介绍 提出图原生强化学习方法，通过概念重组实现可追溯的科学假设生成。将多步推理建模为图上的搜索，每一步结果均可回溯。在材料设计任务中，该方法生成的可验证假设显著优于标准LLM的不可追溯输出。

InstanceControl: Controllable Complex Image Generation without Instance Labeling

👍 8

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，无需实例标注即可实现复杂多实例图像的受控生成。通过隐式实例感知机制解决属性混淆问题，在深度图等条件控制下保持各实例的独立属性，大幅提升多对象场景的生成质量，消除对手动标注的依赖。

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

👍 8

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，一个关注澄清需求的深度搜索基准。评估搜索agent在用户查询不完整或模糊时主动提问澄清的能力。该基准填补了现有评估忽略交互式信息寻求过程的空白，推动更实用搜索agent的发展。

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-预测分离假说：将Transformer中下一token预测和状态存储两种角色解耦。设计一种变体Transformer，其中一部分计算专用于维护隐状态，另一部分专用于预测。实验验证了分离能提升语言建模性能，对架构设计有指导意义。

AnyGroundBench: A Specialized-Domain Benchmark for Video Grounding in Vision-Language Models

👍 7

07/02 08:00

Vision-Language Models (VLMs) have demonstrated immense promise in Spatio-Temporal Video Grounding (STVG). However, current evaluation protocols are largely confined to zero-shot assessments on general, daily-life benchmarks. This creates a critical disconnect from real-world applications in special

中文介绍 提出AnyGroundBench，一个面向专业领域的视频定位基准，用于评估VLM在特殊场景下的时空视频定位能力。填补了通用基准与工业应用之间的鸿沟，覆盖多个专业领域，为VLM的领域适配提供更可靠的评测。

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擅长优化有明确损失函数的问题，而学校教育多为这类问题。未来有价值的工作是无法被损失函数定义的领域，如创造性、不确定性决策。博主给出AI时代的职业建议。

Your AI, your growth

@arthurmensch · 74.0K 粉丝 · 162.7K 阅 · 539 赞 · 76 转

07/04 05:31

Of course you need to use open-source models if you’re an enterprise leader. Close model providers, that are now forcing data retention, are gaining immense leverage on your business if you don’t. As

中文介绍 企业领导者应优先使用开源模型，避免闭源模型提供商强制数据保留而获取业务杠杆。数据自主权是AI应用增长的关键。

How to build a second brain with Fable 5

@EXM7777 · 121.1K 粉丝 · 83.6K 阅 · 518 赞 · 48 转

07/03 22:06

I'm going to show you, step by step, how to turn Fable 5 into a machine that knows your business inside out... and ships outputs that look nothing like what everyone else is getting the tool is a

中文介绍 一步步教程：用Fable 5构建「第二大脑」，让模型深入了解你的业务，输出个性化、与众不同的结果，而非通用答案。

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不是vibe coding工具，而是操作系统级模型。率先用它写代码是错误用法，应探索其调度与编排能力。

Agentic Autonomy Levels

@addyosmani · 404.9K 粉丝 · 53.5K 阅 · 525 赞 · 66 转

07/03 11:29

In most conversations about agentic engineering, the action has changed from prompting to operating. Here's a frontier looking into the fog: software factories, goals, loops, background sessions,

中文介绍 梳理Agent自主等级：从提示工程升级到操作工程。探讨软件工厂、目标循环、后台会话等前沿概念，描绘Agentic系统发展方向。

A Field Guide to Fable: Finding Your Unknowns

@trq212 · 299.3K 粉丝 · 40.1K 阅 · 534 赞 · 43 转

07/04 01:43

Working with Claude Fable 5 keeps re-teaching me an old lesson: the map is not the territory. The map, a representation of the work to be done, is my prompts and skills and context, it’s what I give

中文介绍 使用Claude Fable 5时，提示与技能只是「地图」，而非真正的「疆域」。强调探索未知领域的重要性，而非固守已有认知。

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 Fable回归Cursor，分享使用模式：Fable作为编排者（orchestrator），Composer作为执行工人（workers）。给出充分利用该模型的实用技巧。

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训练管线速查表，覆盖梯度更新、数据加载、损失计算等核心步骤。适合快速回顾或新手学习。

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宣布首次研究合作，具体内容未披露。

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工程师世界博览会闭幕，围绕循环（loops）展开辩论，发布AI工程现状报告，闭幕主旨演讲聚焦下一步构建方向。

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阐述其代理框架“eve”的创建过程，强调技能、沙盒和代理可读网站的重要性。

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

TLDR AI · 07/03 08:00

中文介绍 Meta推出“Watermelon”项目，Anthropic与三星在芯片方面合作，自主研究（autoresearch）进入实践阶段。

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“代理网站”，可根据用户意图动态生成页面。AIEWF上Carlos Sanchez讨论了Web的未来。

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 文章探讨利用AI实现运营卓越。Lean Six Sigma和BPM等框架强调统计严谨性和质量控制，AI正在推动这些方法的演进。

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”项目，强调在循环最大化时代人类判断的重要性，指出代理仍需人类引导。

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在物理基础设施、运营连续性和安全关键行业中的实际应用，超越聊天机器人和图像生成器，正在改变工业场景。

[AINews] not much happened today

Latent Space · 07/02 15:10

another quiet day.

中文介绍 当天AI新闻平淡，无明显重大事件。

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每日简报：自主研究（autoresearch）引发AI与人类自主权之间的张力，多位演讲者捍卫人类理解和控制权。

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将编码代理评估扩展至数据库、API密钥等，实现端到端应用交付。LangChain发布LangSmith（统一追踪）和OpenWiki（自动文档生成）。

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

TLDR AI · 07/02 08:00

中文介绍 Google升级Gemini Flash，Meta推出AI云服务，ZCode项目亮相。

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解释自主研究（autoresearch）、代理“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介绍其前部署工程师团队如何帮助企业实施代理，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 Genesis Molecular AI的Evan Feinberg和Sergey Edunov讨论：Llama负责人离开Meta转向药物发现，PEARL在零样本OpenBind中获胜，共折叠技术突破准确率阈值。