2026-08-23

EnvHarness: Awakening Static Worlds for Agent Learning

👍 252

08/19 20:00

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 提出了一种新的环境生成方法EnvHarness，以促进大规模语言模型（LLM）代理的学习。该方法解决了静态、手工构建环境无法适应代理能力提升的问题，支持自适应环境生成，进而改善代理的交互质量。研究展示了这一方法在提升代理学习效率和适应性上的显著潜力，为动态交互环境的生成提供了新思路。意义：对Agent和自适应学习的研究趋势有重要影响。

Zetta ζ: An Efficient Closed-Loop Embodied Harness for Self-Evolving Physical Intelligence

👍 141

08/16 20:00

Embodied agents are increasingly used to close the gap left by end-to-end policy models. Yet the agentic path has not realized closed-loop learning in physical execution: existing harnesses remain largely open-loop, following fixed skills during rollout and reflecting only after an episode completes

中文介绍 Zetta ζ 提出了一个高效的闭环体现在自我进化的物理智能体平台，旨在弥补现有开放式学习模型的不足。通过闭环学习，这一平台使得代理在物理执行中能实时调整和改进，而非仅依赖于预设的固定技能。研究表明该系统改善了自我学习能力，有助于推动智能体在动态环境中的表现。意义：对 Agent 和自我进化研究有重要推动作用。

SemaPLC: A Project-Grounded, Verification-Gated Agent Harness for PLC Code Generation

👍 115

08/18 20:00

Programmable logic controllers (PLCs) run industrial plants, and large language models can already generate independent program organization units (POUs) for them. Whether such logic integrates into an existing PLC project and then runs correctly has been checked only in limited tests. We present Se

中文介绍 SemaPLC 提出了一个基于项目的验证门控代理架构，以生成可编程逻辑控制器（PLC）代码，确保自动生成的逻辑能够正确集成到现有项目中。通过与实际项目的匹配和验证，该方法提高了生成代码的准确性和可靠性，降低了错误的发生率。研究结果为工业自动化领域中代码生成的应用提供了新工具。意义：推动了智能代码生成和自动化测试的研究进展。

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

👍 114

08/18 20:00

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 FACET 解决了训练高质量终端代理任务中可执行监督的问题，通过合成一组任务，将指令、初始化环境和验证器整合为一致的系统。该方法提升了任务合成的质量，并支持更为系统化和可扩展的训练过程，对任务生成的准确性和可靠性做出了重要贡献。意义：对终端任务生成和增强学习领域具有重要的应用前景。

Co-RL: Unsupervised Reasoning Emerges from Diverse Cohort in Multi-agent RL

👍 92

08/18 20:00

Reinforcement learning (RL) has emerged as a powerful approach for improving reasoning in language and vision-language models, yet its strongest successes still depend heavily on ground-truth supervision (e.g., verifiable reward). Such annotations are costly to obtain and become increasingly scarce

中文介绍 Co-RL 提出了一种多代理强化学习框架，通过多样化群体引导无监督推理的生成。这一方法显著减少了对真实数据监督的依赖，针对推理能力的提升展现出了良好的效果，尤其在推理和语言模型的应用中具有重要的往前推进作用。意义：推动了RLHF及无监督学习的研究发展。

OmniScientist: An Omni-Modal Omni-Discipline AI Scientist

👍 88

08/12 20:00

Recent advances in foundation models have enabled AI scientists to automate increasingly complete research workflows, from hypothesis generation and code execution to manuscript preparation. Yet workflow coverage alone does not provide access to the full evidence on which scientific discovery depend

中文介绍 OmniScientist 在基础模型上增强了人工智能科学家的能力，实现了从假设生成到论文撰写的全自动化研究工作流。尽管工作流覆盖面广，但现有方法未能全面获取科学发现所需的所有证据，研究强调了整合各类知识源的重要性。意义：对科学研究的自动化及其影响力有重要意义。

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

👍 61

08/19 20:00

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 SWE-bench Science 评估了编程代理是否能独立解决科学任务中的工程问题，针对现有评价方法的局限性，强调了以成就为导向的评估方式，提供了更为全面的任务成功率分析。通过这一评估框架，能够更好地指导编程代理在科学应用中的表现。意义：加强了编程代理在科学领域中的应用潜力。

SPADE: Self-Play in Adaptive Synthetic Executable Environments

👍 47

08/18 20:00

Continuous self-improvement requires an ever-expanding pool of self-generated, diverse, adaptive goals. For language agents, existing training environment pools (hand-curated, statically synthesized, or frozen-verifier) keep the goal distribution fixed as the learner scales. We introduce SPADE (Self

中文介绍 SPADE 介绍了一种自我对抗训练框架，促进语言代理在自适应合成执行环境中的表现。该方法通过扩展自我生成的多样化目标池，增强代理的自我改进能力，有效应对了固定目标分布带来的挑战，展现了显著的学习速度提升。意义：有助于推进智能体的自我改进及适应性学习研究。

Training Chemical Plausibility-Aware Large Language Models for Single-Step Retrosynthesis

👍 32

08/18 20:00

Single-step retrosynthesis is a central component of computer-aided synthesis planning, yet its intrinsically one-to-many nature is poorly captured by single-answer evaluation and benchmarking protocols. To address this, we introduce Top-K prompting as a robust training and inference paradigm to bet

中文介绍 针对单步逆合成的问题，提出了化学合理性感知的大型语言模型的训练方法Top-K提示。这种新颖的训练和推理范式能更好地捕获单步逆合成的多样性，提升了预测质量，为化学合成规划的应用提供了一个有效的解决方案。意义：推动了化学创造力和合成策略的智能化发展。

MemTrapBench: Benchmarking Cognitive Traps in LLM Memory Use

👍 31

08/19 20:00

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 MemTrapBench 开展了对大型语言模型内存使用中的认知陷阱的基准评估，针对信息提取、存储和检索的质量提出了一系列新标准。这项研究重点关注如何评估提取信息的实际效果，明确了解大模型内存操作中的潜在问题，对内存技术的优化具有重要启示。意义：对智能体的记忆模型和推理能力的设计具有重要影响。

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

👍 29

08/12 20:00

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 SkillEvo 提出了一个自我再生的进化梯度方案，解决了现有代理技能生成缺乏闭环反馈的困境。通过多回合交互反馈增强，自我优化变得可能，研究表明这一方法显著提升了技能的质量和应用能力。意义：对增强学习和智能体技能演进的研究趋势产生影响。

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

👍 17

08/19 20:00

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 Repo0 设计了一种驱动感知的零到全代码生成方法，挑战了现有大型语言模型在预定义代码库结构下的限制，探讨了直接从自然语言构建完整软件项目的可能性。研究表明，该方法在灵活性和适应性上具有显著优势，为代码生成技术提供了新视角。意义：推动了代码生成领域的创新和实用性。

The More Popular, The Harder to Forget: Adaptive Popularity for LLM Unlearning

👍 17

08/13 20:00

Popular facts are memorised more deeply during pretraining and resist removal longer than rare ones, yet existing LLM unlearning methods apply uniform gradient pressure regardless of training-data frequency. We propose the AdaPop (Adaptive Popularity) method, which combines local token confidence wi

中文介绍 AdaPop 方法关注在大语言模型中高频和低频事实的学习差异，提出了一种自适应权重调整机制，以改善不均匀训练数据下的学习效果。研究表明，这种方法能有效提升低频信息的学习表现，推动了模型对训练周期中信息的适应能力。意义：对模型记忆和知识更新的研究趋势提供新的理解。

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

👍 16

08/20 04:02

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 FlashPrefill V2 改进了大语言模型的长上下文建模能力，通过块稀疏预填充注意力机制降低了计算复杂度，从而加速了处理过程。研究表明，这一机制在保持性能的同时有效减少了计算资源消耗，为长期上下文建模提供了实用的解决方案。意义：对长时间序列建模和大规模语言模型的应用具有积极影响。

Looped Language Models Improve Compositional Tool Calling

👍 16

08/16 20:00

Looped language models have shown promising results on reasoning benchmarks, yet their potential for agentic tool use remains largely unexplored. We study this question in compositional tool-calling settings, where models must coordinate multiple API calls, maintain intermediate state, and preserve

中文介绍 研究了循环语言模型在组合工具调用中的表现，强调了其在协调多API调用和维护中间状态方面的潜力。该研究为提升智能体在复杂任务中的协调能力提供了的实证证据。结果显示，在处理复杂逻辑时循环模型的表现优于传统模型。意义：对工具使用和智能体集成能力的提升具有重要启示。

FM-Bench: A Benchmark for Long-Horizon Management with Competing Agents

👍 16

08/18 20:00

Language model agents now execute bounded tasks reliably. Whether they can sustain effective decision-making over long horizons, where actions have cumulative consequences and the environment responds to their choices, remains largely unmeasured. FM-Bench (Football Management Benchmark) measures thi

中文介绍 FM-Bench 评估了语言模型代理在长期管理中的表现，指出当前模型在长时间跨度内的决策持续性仍需测量和提升。该基准通过提供长时间决策评估框架，促进了对智能体决策能力的全面分析，有助于提升智能体在复杂环境中的应用能力。意义：推动了在复杂决策领域的研究进展。

Decision-Metric Alignment in Latent World Models: Diagnostics and Action-Conditioned Objectives for MPC Planning

👍 15

08/18 20:00

JEPA-style latent world models can use Euclidean distance to a goal latent as the cost for model-predictive control (MPC). Strong decoding of task variables, however, does not guarantee that this particular cost ranks candidate action sequences by real task progress. We call the latter property deci

中文介绍 该研究提出了决策度量对齐的概念，分析了JEPA风格潜在世界模型在模型预测控制中的应用。研究强调通过有效的回馈机制来识别真实任务进展所需的动作序列，从而促进代理在动态环境中的表现优化。意义：对潜在世界模型及其在控制策略中的应用具重要理论启示。

Scaling Creative Writing Beyond Story-Centric Data with Attribute-Guided Genre Expansion

👍 14

08/13 20:00

High-quality creative writing data for large language models (LLMs) remains dominated by story-centric data, limiting models' ability to follow the structural and functional conventions of diverse creative formats. We propose an attribute-guided genre expansion framework for scaling creative writing

中文介绍 该研究提出了一种基于属性指导的体裁扩展框架，旨在提升大语言模型生成的创意写作样本的多样性。通过打破传统的故事中心数据限制，支持创新性内容的生成。这一方法为丰富创作类型提供了新的方向，为创意写作教学和实践提供了新的视角。意义：提升了创意写作与多样性研究的前景。

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

👍 13

08/18 09:09

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 该研究分析了大规模混合专家模型在低资源语言推理中的表现，强调了标准精度基准可能掩盖模型实际修复能力的风险。结果提示，量化模型在特定任务下的能力需要更深入的探索，为低资源语言处理研究提供了启发。意义：对低资源语言生成功能的增强提供了新洞见。

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

👍 11

08/18 20:00

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 VA-Judger 提出了从人类偏好反馈中进行奖励建模的方法，旨在增强视频音频协同生成模型的效果。通过综合多质量维度的评估指标，该方法提升了生成内容的整体协调性和质量，展示了在多模态生成领域中的高效应用潜力。意义：对视频生成与多模态学习的深入研究具有推动作用。

AI Engineering Skills Map: Building and Deploying AI Applications

@AndrewYNg · 1.8M 粉丝 · 62.4K 阅 · 1.2K 赞 · 189 转

08/21 12:37

I previously wrote about our AI Engineering Skills Map, with the highest level skills being (i) Building and deploying AI applications, (ii) Software engineering fundamentals, (iii) Using coding

中文介绍 博主分享了人工智能工程技能地图，强调了构建和部署AI应用、软件工程基础以及编程技能等核心能力。这一地图旨在指导未来的AI工程师，反映出行业对全面技能的需求日益增长。

RL Policy Churn

@ID_AA_Carmack · 3.7M 粉丝 · 58.8K 阅 · 577 赞 · 40 转

08/20 15:01

I have a good theory about this now. The Phenomenon of Policy Churn https://arxiv.org/pdf/2206.00730 The classic exploration method in RL is “epsilon greedy”, where the agent normally takes the “greedy” action

中文介绍 博主探讨了强化学习中政策变动的现象，提出了关于“epsilon greedy”探索方法的理论。这表明在强化学习应用中，政策的持续变化是一个重要的研究方向，行业对此现象的理解正在加深。

Grok Bot vs. Hermes: They Look Similar. Underneath, They’re Making Very Different Bets.

@themahis · 4.7K 粉丝 · 43.0K 阅 · 515 赞 · 37 转

08/20 11:59

One is trying to make AI teammates feel like a managed product. The other is turning the agent itself into an open, composable runtime. Research note — August 20, 2026: For this comparison, I

中文介绍 博主比较了Grok Bot和Hermes两种AI代理的设计理念，前者将AI设置为管理产品，后者则强调开放式和可组合的运行时。这显示了AI代理在市场定位和技术开发上存在的不同策略，反映了行业对灵活性和效能的持续探索。

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

Latent Space · 08/22 03:36

Did you think RSI stopped at model training?

中文介绍 研究指出，在模拟技术的快速发展中，模型性能可能减少10%，但成本却下降至原来的1/100，速度提高至原来的10000倍。

The Evolution of the Agent Harness

Latent Space · 08/22 03:30

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 模型正在将代理整合进其权重中，未来将成为关注人类的工具，而非仅仅为模型服务。

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

Latent Space · 08/21 19:37

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 Simile AI的CEO提到，从生成代理到创建80亿个数字双胞胎的过程，探讨了技术转向严肃商业的原因。

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

DeepMind Blog · 08/21 07:59

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

中文介绍 谷歌DeepMind与游戏工作室合作，旨在原型开发突破性的AI游戏玩法，积累了15年的游戏AI研究经验。

[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 01:45

Yes, we’re confused too.

中文介绍 Poolside获得120亿美元的逆向招聘协议，创始人持有10亿美元，员工获得60亿美元，Infraco的云计算能力扩展至7GW。

not much happened today

Smol AI News · 08/21 01:44

**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 Ox Alpha新模型展现出强大的编码能力，分析师认为其性能提升源于后期训练和基础设施改善，而非单纯的模型规模。

Measuring benchmark optimization in speech recognition

Hugging Face Blog · 08/20 20:00

中文介绍 本条目未提供具体内容，摘要无法生成。

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

TLDR AI · 08/20 20:00

中文介绍 本条目未提供具体内容，摘要无法生成。

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

Latent Space · 08/20 16:59

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介绍了他的/wayfinder技能，用于绿地项目或方向不明确时的导航。

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

Hugging Face Blog · 08/20 12:52

中文介绍 本条目未提供具体内容，摘要无法生成。

Debates over AI consciousness are a trap

MIT Tech Review AI · 08/20 11:42

“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”

中文介绍 关于AI意识的辩论被视为误区，影响行业的领导者如Demis Hassabis、Dario Amodei和Sam Altman呼吁对此进行监管。

Unlocking hidden revenue streams with market models

MIT Tech Review AI · 08/20 05:47

Each day, an airline transports tens of thousands of passengers on hundreds of flights. Often these are not straightforward point-to-point routes, with passengers requiring multiple connections. The airline can consider potentially hundreds of variables to price each of these journeys: demand, seaso

中文介绍 航空公司在运输成千上万的乘客时，需要考虑多个变量，以制定合理的票价策略，涉及复杂的市场模型。

Introducing AI Futures

OpenAI News · 08/20 03:00

Introducing AI Futures, a new OpenAI blog exploring how transformative AI could reshape power, governance, the economy, and individual freedom.

中文介绍 OpenAI发布新博客《AI Futures》，探讨变革性AI如何重塑权力、治理、经济和个人自由。

not much happened today

Smol AI News · 08/20 01:44

**OpenAI** and **Anthropic** expanded their agent platforms with new desktop features, collaborative editing, and composable APIs like Skills and Files API. **OpenAI** rolled out memory and workflow features in the EEA, UK, and Switzerland. **AT&T** revealed that 40% of employee AI usage routes to o

中文介绍 OpenAI和Anthropic扩展了其代理平台，新增桌面功能和可组合的API。OpenAI在欧盟、英国和瑞士推出了内存和工作流功能。

[AINews] Death of Params: Z.ai CEO Jie Tang on GLM 5.3 and the new Post-training Scaling Law

Latent Space · 08/20 01:17

Every lab CEO is on X now

中文介绍 Z.ai的首席执行官Jie Tang谈到GLM 5.3和新的后期训练扩展法，指出该领域的实验室CEO都正在X平台上活跃。