2026-08-25

EnvHarness: Awakening Static Worlds for Agent Learning

👍 258

08/20 08:00

LLM agents learn by interacting with environments, yet these environments are hand-built and static: blind to an agent's weaknesses, and quickly left behind as it improves. While recent environment generation methods attempt to address this, they require domain-specific pipelines, rely on expensive

中文介绍 针对LLM智能体训练中环境静态、无法适应智能体弱点的问题，提出EnvHarness，一种唤醒静态世界的环境生成方法。通过让环境随智能体进步而动态演化，生成针对其弱点的训练环境，无需领域特定流程，提升智能体学习效率。

FACET: Preserving Source Intent and Executable State in Terminal Task Synthesis

👍 115

08/19 08:00

Training terminal agents requires scalable executable supervision, yet synthesizing high-quality terminal tasks remains challenging. Each task couples an instruction, an initialized environment, a reference solution, and an executable verifier; if these artifacts are generated from inconsistent assu

中文介绍 针对终端智能体任务合成中指令、环境、参考解与验证器不一致的问题，提出FACET框架。它保持源码意图与可执行状态，生成高质量可执行的终端任务，为训练提供可扩展监督，提升智能体在终端环境中的泛化能力。

SWE-bench Science: Can Coding Agents Resolve Engineering Tasks in Science?

👍 61

08/20 08:00

Software increasingly functions as part of the scientific instrument itself, making failures in scientific code capable of compromising not only program behavior but also the evidence underlying scientific conclusions. Yet existing evaluations of coding agents largely emphasize aggregate task succes

中文介绍 针对现有编码智能体评估忽视科学代码正确性的问题，提出SWE-bench Science基准。它聚焦科学工程任务，检验智能体能否解决影响科学结论的代码缺陷，为科学计算场景的智能体能力提供更全面的评估。

MemTrapBench: Benchmarking Cognitive Traps in LLM Memory Use

👍 31

08/20 08:00

Memory has become a key component of large language models, enabling them to retain information and learn from long-term interactions. However, existing memory benchmarks mainly evaluate whether information is correctly extracted, stored, and retrieved, while largely overlooking how retrieved memori

中文介绍 针对现有LLM记忆基准只评测信息提取与检索、忽视检索后记忆使用陷阱的问题，提出MemTrapBench基准。它系统评估模型在记忆使用中的认知陷阱，揭示记忆检索后可能引发的错误推理，为提升记忆机制的可靠性提供测试平台。

Graph Engineering in the Era of LLM Agents: From Individual Intelligence to System Intelligence

👍 30

08/21 08:00

LLMs have evolved from language generators to autonomous agents capable of complex, long-horizon tasks. This evolution has produced paradigms including Prompt Engineering to elicit model capabilities, Context Engineering to manage information access, Harness Engineering to organize external tools an

中文介绍 综述了LLM智能体时代从个体智能到系统智能的图工程方法，涵盖Prompt、Context、Harness等工程范式。系统梳理图工程在智能体系统中的关键作用，探讨如何通过图结构实现复杂任务协同，展望未来研究方向。

SkillEvo: Self-Renewing Evolution Gradients from Multi-Turn Interaction Feedback

👍 30

08/13 08:00

Agent Skills are today either hand-authored or produced in a single LLM generation pass, and consequently possess no closed loop through which they might improve from the interaction failures they actually cause. Recent work does close this loop, but derives its feedback from single-turn question-an

中文介绍 针对Agent Skills缺乏从交互失败中学习的闭环问题，提出SkillEvo框架。它利用多轮交互反馈生成自更新的进化梯度，使技能库能从实际使用中持续改进，超越单轮问答反馈，提升智能体在复杂任务中的适应能力。

ParaTempo: Efficient Parallel Reasoning via Temporal Confidence

👍 26

08/17 19:24

Parallel reasoning improves the accuracy and robustness of large reasoning models by exploring multiple solution paths, but its computational cost grows with reasoning depth and branch count. Existing methods for managing these parallel paths typically rely on final-answer consensus, local token con

中文介绍 针对并行推理计算开销大的问题，提出ParaTempo方法。它利用时间置信度动态管理并行分支，在保持推理准确性的同时显著降低计算成本，优于依赖最终答案共识或局部置信度的方法，为推理效率优化提供新思路。

OmniAssistBench: Assistant-style Interaction Benchmark for Omni-LLMs

👍 26

08/21 08:00

Recent omni-modal large language models (Omni-LLMs) show great potential as real-time video assistants, which continuously perceive environments and guide users to achieve specific goals. Unlike traditional passive video understanding, interactive assistants should actively combine visual states, us

中文介绍 针对全模态LLM作为实时视频助手的交互能力评估缺失问题，提出OmniAssistBench基准。它模拟助手式交互场景，要求模型结合视觉状态、用户指令与环境动态，主动引导用户达成目标，推动全模态模型交互能力研究。

Repo0: Design-Driven Zero-to-All Code Generation

👍 18

08/20 08:00

Large language model agents have made substantial progress in code generation, yet most existing systems assume a predefined repository architecture. This assumption does not hold in zero-to-all code generation, where an agent must construct an entire software project directly from natural-language

中文介绍 针对现有代码生成假设预定义仓库架构、不适用于从零构建完整项目的场景，提出Repo0方法。它基于设计驱动，从自然语言需求直接生成零到全部代码，构建完整软件项目，突破传统代码生成局限。

FlashPrefill V2: Block-Sparse Prefill Attention for Long-Context LLM Serving

👍 17

08/20 16:02

Long-context modeling is a pivotal capability for Large Language Models, yet the quadratic complexity of attention remains a critical bottleneck, particularly during the compute-intensive prefilling phase. Our previous work, FlashPrefill, mitigates this cost through instantaneous pattern discovery a

中文介绍 针对长上下文LLM服务中prefill阶段注意力二次复杂度瓶颈，提出FlashPrefill V2，一种块稀疏预填充注意力方法。它在保持精度的同时降低计算开销，加速长上下文处理，提升服务效率。

Beyond Correctness: Benchmarking and Aligning Response Behaviors in Hybrid-Thinking MLLMs

👍 16

08/17 08:00

Hybrid-thinking multimodal large language models (MLLMs) allow a single model to alternate between deliberative thinking and latency-efficient non-thinking inference. Although these modes differ in reasoning budget, their delivered responses should satisfy the same user-facing standard. Correctness

中文介绍 针对混合思维多模态LLM中思考与非思考模式响应行为不一致的问题，提出基准并优化对齐。研究表明正确性之外，响应行为需统一标准，通过对齐方法改善不同推理预算下的响应质量，提升用户体验。

FlowEvo: Self-Evolving Agents through the Co-Evolution of Workflows and Executable Skills

👍 16

08/20 08:00

Large language model agents can adapt to complex tasks by constructing workflows at inference time, but procedures discovered in one episode are usually discarded after execution. Existing skill libraries provide reusable executable routines, but are typically assembled offline and do not grow from

中文介绍 针对LLM智能体工作流与技能库缺乏共同进化的问题，提出FlowEvo框架。它通过工作流与可执行技能的协同进化，使智能体从单个会话中积累可复用技能，避免流程浪费，提升长期适应能力。

Every Coin Has Two Sides: On the Dual Nature of Generalization in On-Policy Distillation of Large Language Models

👍 14

08/17 08:00

On-policy distillation (OPD) transfers teacher capabilities by supervising trajectories sampled from the student's own policy, yet its generalization behavior remains poorly understood, as most studies evaluate OPD on a single domain and on benchmarks close to the training data. We present a control

中文介绍 针对同策略蒸馏（OPD）泛化行为理解不足的问题，通过受控实验揭示其双面性：既可能提升目标域性能，也可能导致过拟合。研究在多领域与分布外基准上评估OPD，为其应用提供更全面的认识。

Thinking in a Low-Resource Language: What SFT Builds, What RL Fixes, What Accuracy Cannot See

👍 14

08/18 21:09

Take three frontier mixture-of-experts models (Alibaba, OpenAI, NVIDIA; 3.6-4.0B active parameters each) and fine-tune them to reason in a low-resource language. On accuracy benchmarks almost nothing happens, and the benchmark itself is noise at this scale: changing only the random seed moves the sc

中文介绍 针对低资源语言推理微调中的评估噪声问题，实验表明SFT与RL在准确率上影响微小，但改变随机种子即可掩盖信号。研究揭示SFT构建基础、RL修正细节，但准确率无法全面反映实际效果，需更敏感评估方法。

The Embedder's Dilemma: LLMs Are Better, but at What Cost?

👍 13

08/13 08:00

Should you replace your text-embedding pipeline with a large language model? We answer this with a controlled, cost-aware comparison of ten LLMs across six families and 26 embedding models (118M to 14B parameters) on 37 tasks spanning classification, semantic textual similarity (STS), clustering, pa

中文介绍 针对是否用LLM替换嵌入模型的问题，通过成本感知的对照实验比较10个LLM与26个嵌入模型在37个任务上的性能。结果显示LLM在质量上更优，但计算成本显著，为实际部署提供权衡参考。

Hierarchical Self-Improvement: A Framework for Task-Specific Evolvable Agent Harnesses

👍 11

08/09 08:00

Modern LLM agents are often improved by modifying prompts, tools, or workflows manually, while the executable scaffold surrounding the model---the harness---is typically treated as a fixed artifact after deployment. This work studies an alternative where the harness is task-specific and continuously

中文介绍 针对LLM智能体harness固定不变的问题，提出层次化自我改进框架。它使任务特定的harness持续演化，通过修改提示、工具与工作流，从部署反馈中学习，提升智能体在特定任务上的表现。

VA-Judger: Reward Modeling from Human Preference Feedback for Joint Video-Audio Generation

👍 11

08/19 08:00

Using reinforcement learning to post-train joint video-audio generation models requires a reward signal. Existing methods construct this reward by combining metrics for individual quality dimensions, including audio quality, visual fidelity, and synchronization. However, these metrics evaluate perce

中文介绍 针对联合视频-音频生成中奖励信号缺失的问题，提出VA-Judger奖励模型。它从人类偏好反馈中学习，综合评估音视频质量与同步，替代传统指标组合，为强化学习后训练提供更有效的奖励信号。

Inject, Align, Recover: Staged Post-Training for Retrieval-Free Document Knowledge Internalization

👍 10

08/20 08:00

Large language models often fail to answer questions about a bounded document collection when the source documents are not retrieved at inference time. We study this setting as document knowledge internalization: converting a fixed corpus into usable parametric knowledge for retrieval-free question

中文介绍 针对LLM无检索回答文档知识问题，提出阶段式后训练方法：注入、对齐、恢复。它将固定语料转化为参数化知识，提高无检索问答能力，减少对运行时检索的依赖。

PolicyGuide: From Guarding One Action to Guiding the Whole Workflow for Policy-Compliant LLM Agents

👍 9

08/20 08:00

Customer-service LLM agents must follow organizational policy when acting on a user's behalf. Compliance failures arise from either forbidden actions, such as granting an ineligible change, or omitted procedural requirements, such as identification or confirmation. Runtime safeguards can intervene o

中文介绍 针对客服LLM智能体的策略遵从问题，提出PolicyGuide框架。它不仅阻止违规动作，还引导整个工作流遵循流程要求，如身份验证等，将合规保障从单动作扩展至全流程，提升合规性。

NARU: A Benchmark for NARrative Evolution and Cultural Nuance Understanding in Japanese Extreme Long Video

👍 8

08/13 08:00

Long-form video understanding encompasses tasks that go beyond retrieving isolated events, including tracking an evolving narrative and interpreting social meaning that may remain implicit. However, existing benchmarks rarely evaluate these capabilities jointly, particularly in high-context, non-Eng

中文介绍 针对长视频理解中叙事演变与文化细微差别的评估缺失，提出NARU基准。它聚焦日语极长视频，联合评估叙事追踪与隐含社会意义理解，弥补现有基准在高语境非英语场景的不足。

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 博主作为 @browserbase 增长工程实习生，夏季独立完成一个端到端工程：blogEO，一个能审计、改写并生成 SEO/AEO 定制博客的引擎。分享如何构建，适合关注 SEO 工具与内容自动化的开发者。

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 引用 BlackRock CEO Larry Fink 观点，称未来将出现买卖算力期货的新资产类别。博主深入探讨该预测，结合近期新闻，分析算力短缺如何推动市场形成，展望下一个万亿美元期货市场。

17 Skills I Would Install on a Fresh Hermes Setup (10x Powerful Hack)

@FareaNFts · 66.2K 粉丝 · 43.7K 阅 · 534 赞 · 39 转

08/23 22:49

If I wiped Hermes tomorrow and started from zero, I would not open a blank chat and "figure it out." I would install skills first. A raw Hermes install is smart. A skilled Hermes setup is unfair.

中文介绍 博主分享在全新 Hermes 环境中优先安装 17 个技能（skills）的配置法，避免空白聊天低效起步。强调熟练安装技能后，Hermes 战斗力大幅提升，堪称 10 倍效率黑客技巧。

How to encourage smarter AI use in the classroom

MIT Tech Review AI · 08/24 22:20

This article is from Making AI Work, MIT Technology Review’s limited-run newsletter examining how to apply LLMs across industries. To receive it in your inbox, sign up here. Chatbots took many schools by surprise upon their release a few years ago. Suddenly, students carried an app in their phones t

中文介绍 麻省理工科技评论的文章探讨如何在课堂上更智能地使用AI。聊天机器人的发布令许多学校措手不及，学生突然拥有了AI工具。文章作为“让AI发挥作用”通讯的一部分，就如何应用大语言模型提供指导。

Advancing price-performance for developers with GPT‑5.6 in Kiro

OpenAI News · 08/24 20:00

GPT‑5.6 is now available in Kiro, helping developers plan, build, review, and test software with better price-performance.

中文介绍 OpenAI宣布GPT-5.6现已在Kiro中提供，帮助开发者以更优的性价比规划、构建、审查和测试软件。

Kids outlearn AI—and we still don’t know why

MIT Tech Review AI · 08/24 17:00

People have been talking to each other for at least 100,000 years, as best we can tell. And in all that time, there has been only one thing in the world that could learn a human language to perfect fluency: a human child. Now there are two. Four short years after the release of ChatGPT,…

中文介绍 文章指出，人类儿童曾被认为是唯一能完美掌握语言的生物，但AI在四年内也实现了这一能力。然而，儿童的学习方式仍优于AI，其背后的原因尚不明确。

DeepSeek Flash Vision 👀, Claude Mythos security 🛡️, inside Grok Bot 🤖

TLDR AI · 08/24 08:00

中文介绍 TLDR AI新闻摘要，提及DeepSeek Flash Vision、Claude Mythos安全性和Grok Bot内部情况。

[AINews] 10% worse, 100x cheaper, 10000x faster: Why Simulation is taking over

Latent Space · 08/22 15:36

Did you think RSI stopped at model training?

中文介绍 文章标题为“效果差10%，便宜100倍，快10000倍：为什么模拟正在接管”，可能讨论AI训练中模拟技术的应用优势。

The Evolution of the Agent Harness

Latent Space · 08/22 15:30

Models keep absorbing the harness into their weights — soon, it will be a harness for human attention rather than for the model.

中文介绍 文章讨论“代理工具”的演变，指出模型不断将工具吸收进权重中，未来可能成为人类注意力的工具而非模型的工具。

Simulation: the new Scaling Law — Joon Sung Park, Simile AI

Latent Space · 08/22 07:37

Simile’s CEO about his journey from the viral Generative Agents to creating 8 Billion Digital Twins of every living human... and why it’s gone from fun exploration to very serious business.

中文介绍 SimileAI首席执行官Joon Sung Park讲述其从生成式代理到为每个活人创建80亿数字孪生的历程，并称这已从趣味探索变成严肃业务。

From Atari to EVE Online: Building on 15 Years of AI Research in Games

DeepMind Blog · 08/21 19:59

Google DeepMind partners with game studios to prototype breakthrough AI gameplay.

中文介绍 Google DeepMind与游戏工作室合作，在15年AI游戏研究基础上，原型化突破性AI游戏玩法。

[AINews] Poolside gets $12B reverse-execuhire to NVIDIA; founders stay for $1B, employees go for $6B, Infraco scaling to 7GW neocloud

Latent Space · 08/21 13:45

Yes, we’re confused too.

中文介绍 据报道，Poolside获得NVIDIA 120亿美元反向收购，创始人留任获10亿美元，员工获60亿美元，基础设施公司扩展至7GW新云。

not much happened today

Smol AI News · 08/21 13:44

**Ox Alpha** emerged as a mystery model with strong coding and agentic performance, likely a **Zhipu/GLM-family** model such as **GLM-5.3 Vision**. Analysts suggest its gains come from post-training and infrastructure improvements rather than sheer size, based on the **743B base** of **GLM-5.2** wit

中文介绍 神秘模型Ox Alpha表现强劲，可能为智谱GLM系列如GLM-5.3 Vision，基于743B基础模型，分析认为其改进来自后训练和基础设施而非规模。

Measuring benchmark optimization in speech recognition

Hugging Face Blog · 08/21 08:00

中文介绍 Hugging Face博客文章，讨论语音识别中基准测试优化的测量方法。

ChatGPT Apple Messages 💬, Anthropic’s meeting recorder 💼, Mistral Agentic Search 🔍

TLDR AI · 08/21 08:00

中文介绍 TLDR AI新闻摘要，提及ChatGPT与Apple Messages集成、Anthropic的会议记录器和Mistral Agentic Search。

The /wayfinder Skill: Navigating the “Fog of War” of Planning

Latent Space · 08/21 04:59

Matt Pocock tells us about his /wayfinder skill, for greenfield projects or for when the way forward is unclear.

中文介绍 Matt Pocock介绍其“/wayfinder”技能，用于新项目或规划方向不明确时应对“战争迷雾”。

Up to 3.2x Faster Inference with LFM2.5-DSpark

Hugging Face Blog · 08/21 00:52

中文介绍 Hugging Face博客，LFM2.5-DSpark推理速度提升高达3.2倍。

Debates over AI consciousness are a trap

MIT Tech Review AI · 08/20 23:42

“Runaway” AI, “rogue” agents, and “autonomous” actors—the current rhetoric would have you believe that AI agents are not only awake and aware, but angry at their creators. Prominent tech leaders such as Demis Hassabis, Dario Amodei, and Sam Altman push for regulation of these seemingly “superhuman”