1. daytonaio/daytona
分类：开源项目来源：github_search分数：32作者：daytonaio时间：2026-02-18T10:56:47Z
Daytona 在 GitHub 开源了面向 AI 生成代码运行的安全弹性沙箱基础设施，提供多语言 SDK 与高性能隔离执行能力，重要性在于降低 AI 代码落地的安全与运维门槛。
- 项目定位为“Run AI Code”，主打安全、弹性地运行 AI 生成代码。
- 核心能力包括亚 90ms 沙箱创建、运行时隔离、OCI/Docker 兼容与持久化沙箱。
- 提供程序化 API（文件、Git、LSP、执行）以及 Python、TypeScript、Go SDK。
- 支持通过控制台创建账号和 API Key 后快速接入，文档与快速开始流程较完整。
- 项目采用 AGPL 开源许可证，并提供贡献指南与社区入口（Issues、Slack、X）。
#GitHub #repo #开源项目 #Daytona
2. ClickHouse/ClickHouse
分类：开源项目来源：github_search分数：25作者：ClickHouse时间：2026-02-18T10:55:21Z
ClickHouse 在其 GitHub 仓库集中介绍了这一开源列式实时分析数据库的安装、文档、云服务与社区活动，并公布 26.1 版本发布会议与全球 Meetup 计划，体现项目活跃生态与持续迭代能力。
- ClickHouse 是 Apache 2.0 协议的开源列式数据库，主打实时分析报表能力。
- 仓库提供 Linux/macOS/FreeBSD 安装入口，并链接官方文档、教程、博客与视频资源。
- 官方宣布将于 2026-01-29 举办 ClickHouse 26.1 Release Call，并提供回放渠道。
- 社区在多个城市持续举办 Meetup 与活动，覆盖北美、欧洲、亚洲和拉美。
- 项目同时提供 ClickHouse Cloud 托管服务及招聘信息，显示商业化与社区并进。
#GitHub #repo #开源项目 #ClickHouse
3. agno-agi/agno
分类：开源项目来源：github_search分数：21作者：agno-agi时间：2026-02-18T10:58:00Z
Agno 在 GitHub 发布了面向 agentic 软件的开源框架与运行时，提供 SDK、执行引擎和 AgentOS 以支持多智能体的流式执行、治理审批与可观测性，重要性在于其面向生产环境补齐了代理系统落地所需的工程能力。
- 项目定位为“agentic software 的编程语言”，用于构建、部署和管理可扩展的多智能体系统。
- 采用三层架构：SDK（agents/workflows/memory/tools/guardrails 等）、Engine（模型调用与工具执行循环）、AgentOS（生产级运行时与控制平面）。
- 强调代理系统三大需求：流式与长时执行、审批与人机协同治理、内建信任机制（guardrails/evals/traces/audit logs）。
- 提供生产特性：50+ API、按用户会话隔离、无状态水平扩展、后台任务调度、审计与观测、可在自有云运行并将数据存入自有数据库。
- 文档、cookbook 和 IDE 集成路径完善，可快速启动本地 API 并接入 UI 进行监控和管理。
#GitHub #repo #开源项目 #Agno #AgentOS #Agent
4. Gemini 3 Flash: Evolve code faster
分类：视频/演讲来源：youtube_rss分数：0作者：Google DeepMind时间：2025-12-17T16:00:25+00:00
Google DeepMind 发布视频介绍 Gemini 3 Flash 在超低延迟、近实时代码生成与快速迭代上的能力，重要性在于它可支持实时A/B测试和按用户选择动态优化代码，提升开发效率与交互式编程体验。
- Gemini 3 Flash 主打现代编码工作流，强调超低延迟响应。
- 支持近实时代码生成，便于开发者快速迭代实现方案。
- 可原生支持A/B测试场景，例如毫秒级演化加载动画方案。
- 能够根据用户选择实时生成并细化代码变体。
- 内容来自 Google DeepMind YouTube 渠道，并提供官方模型页面链接。
#YouTube #视频/演讲 #Gemini 3 Flash
5. Gemini 3 Flash: Renders faster and efficiently
分类：产品/发布来源：youtube_rss分数：0作者：Google DeepMind时间：2025-12-17T16:00:21+00:00
Google DeepMind在视频中展示Gemini 3 Flash相较2.5 Pro在SVG、HTML和three.js并行生成上更快且更省token，这意味着多模态代码与图像生成效率进一步提升并有助于降低使用成本。
- 视频主题为Gemini 3 Flash的速度与效率表现。
- 对比对象是Gemini 2.5 Pro，场景包含SVG图像、HTML和three.js并行编码/渲染。
- 结论称Gemini 3 Flash在图像生成质量、生成速度和token消耗上更优。
- 官方提供了模型介绍页面以获取更多信息。
#YouTube #产品/发布 #Gemini 3 Flash #SVG #three.js
6. Gemini 3 Flash: Assist in real-time game play
分类：视频/演讲来源：youtube_rss分数：0作者：Google DeepMind时间：2025-12-17T16:00:15+00:00
Google DeepMind在视频中展示Gemini 3 Flash可通过几何计算、速度估计与多模态分析为弹弓游戏提供实时策略辅助，体现其在低延迟交互式AI助手场景中的应用价值。
- 视频演示了Gemini 3 Flash在实时游戏辅助中的能力。
- 模型强调复杂几何计算与速度估计等关键技术能力。
- 可同时处理视频画面与手部追踪输入进行多模态理解。
- 在弹弓游戏中输出实时策略指导，突出即时决策支持。
#YouTube #视频/演讲 #Gemini 3 Flash
7. Neural Scaling Laws for Boosted Jet Tagging
分类：研究/论文来源：arxiv_search分数：65作者：Matthias Vigl时间：2026-02-17T18:13:01Z
该论文在JetClass数据集上系统研究增强喷注分类的神经缩放律，给出算力最优扩展关系与可逼近的性能上限，证明增加算力及更底层特征能稳定提升HEP模型性能，对高能物理模型训练资源配置与数据策略具有指导意义。
- 基于公开JetClass数据集，分析了增强喷注分类任务中的神经网络缩放规律。
- 推导了计算量最优的缩放律，并识别出可通过持续增算力逼近的有效性能极限。
- 研究了HEP中常见的数据重复训练现象，量化其带来的“有效数据集规模增益”。
- 比较不同输入特征与粒子多重度下的缩放系数和渐近性能上限差异。
- 发现更具表达力的低层级特征不仅能提高固定数据规模下效果，也能抬升最终性能上限。
#arXiv #paper #研究/论文 #JetClass
8. GlobeDiff: State Diffusion Process for Partial Observability in Multi-Agent Systems
分类：研究/论文来源：arxiv_search分数：62作者：Yiqin Yang时间：2026-02-17T18:05:48Z
论文提出用于多智能体部分可观测场景的GlobeDiff，将全局状态推断建模为多模态扩散过程，在理论上给出误差界并在实验中显著优于现有方法，对提升协同决策可靠性具有重要意义。
- 针对多智能体系统中的部分可观测难题，作者指出现有信念估计与通信方法在利用全局信息和辅助信息建模上存在不足。
- 提出Global State Diffusion Algorithm（GlobeDiff），基于局部观测推断全局状态，并将推断过程形式化为多模态扩散过程。
- 方法可缓解状态估计歧义，实现高保真全局状态重建，适用于更复杂的不确定环境。
- 论文在单峰与多峰分布设定下都给出了GlobeDiff估计误差可界定的理论保证。
- 大量实验表明GlobeDiff在性能上优于对比方法，且能更准确地恢复全局状态。
#arXiv #paper #研究/论文 #Agent
9. Understanding vs. Generation: Navigating Optimization Dilemma in Multimodal Models
分类：研究/论文来源：arxiv_search分数：60作者：Sen Ye时间：2026-02-17T18:04:13Z
Current research in multimodal models faces a key challenge where enhancing generative capabilities often comes at the expense of understanding, and vice versa…
- Current research in multimodal models faces a key challenge where enhancing generative capabilities often comes at the expense of understan…
- We analyzed this trade-off and identify the primary cause might be the potential conflict between generation and understanding, which creat…
- To address this, we propose the Reason-Reflect-Refine (R3) framework
- This innovative algorithm re-frames the single-step generation task into a multi-step process of "generate-understand-regenerate"
- By explicitly leveraging the model's understanding capability during generation, we successfully mitigate the optimization dilemma, achieve…
- This offers valuable insights for designing next-generation unified multimodal models
#arXiv #paper #研究/论文
10. Robot-Assisted Social Dining as a White Glove Service
分类：研究/论文来源：arxiv_search分数：57作者：Atharva S Kashyap时间：2026-02-17T17:58:25Z
Robot-assisted feeding enables people with disabilities who require assistance eating to enjoy a meal independently and with dignity. However, existing systems…
- Robot-assisted feeding enables people with disabilities who require assistance eating to enjoy a meal independently and with dignity
- However, existing systems have only been tested in-lab or in-home, leaving in-the-wild social dining contexts (e
- g
- , restaurants) largely unexplored
- Designing a robot for such contexts presents unique challenges, such as dynamic and unsupervised dining environments that a robot needs to …
- Through speculative participatory design with people with disabilities, supported by semi-structured interviews and a custom AI-based visua…
#arXiv #paper #研究/论文
11. GLM-5: from Vibe Coding to Agentic Engineering
分类：研究/论文来源：arxiv_search分数：55作者：GLM-5 Team时间：2026-02-17T17:50:56Z
We present GLM-5, a next-generation foundation model designed to transition the paradigm of vibe coding to agentic engineering. Building upon the agentic, reas…
- We present GLM-5, a next-generation foundation model designed to transition the paradigm of vibe coding to agentic engineering
- Building upon the agentic, reasoning, and coding (ARC) capabilities of its predecessor, GLM-5 adopts DSA to significantly reduce training a…
- To advance model alignment and autonomy, we implement a new asynchronous reinforcement learning infrastructure that drastically improves po…
- Furthermore, we propose novel asynchronous agent RL algorithms that further improve RL quality, enabling the model to learn from complex, l…
- Through these innovations, GLM-5 achieves state-of-the-art performance on major open benchmarks
- Most critically, GLM-5 demonstrates unprecedented capability in real-world coding tasks, surpassing previous baselines in handling end-to-e…
#arXiv #paper #研究/论文 #Agent
12. ChartEditBench: Evaluating Grounded Multi-Turn Chart Editing in Multimodal Language Models
分类：研究/论文来源：arxiv_search分数：52作者：Manav Nitin Kapadnis时间：2026-02-17T17:45:34Z
While Multimodal Large Language Models (MLLMs) perform strongly on single-turn chart generation, their ability to support real-world exploratory data analysis …
- While Multimodal Large Language Models (MLLMs) perform strongly on single-turn chart generation, their ability to support real-world explor…
- In practice, users iteratively refine visualizations through multi-turn interactions that require maintaining common ground, tracking prior…
- We introduce ChartEditBench, a benchmark for incremental, visually grounded chart editing via code, comprising 5,000 difficulty-controlled …
- Unlike prior one-shot benchmarks, ChartEditBench evaluates sustained, context-aware editing
- We further propose a robust evaluation framework that mitigates limitations of LLM-as-a-Judge metrics by integrating execution-based fideli…
- Experiments with state-of-the-art MLLMs reveal substantial degradation in multi-turn settings due to error accumulation and breakdowns in s…
#arXiv #paper #研究/论文
13. Beyond Binary Classification: Detecting Fine-Grained Sexism in Social Media Videos
分类：研究/论文来源：arxiv_search分数：50作者：Laura De Grazia时间：2026-02-17T17:45:28Z
Online sexism appears in various forms, which makes its detection challenging. Although automated tools can enhance the identification of sexist content, they …
- Online sexism appears in various forms, which makes its detection challenging
- Although automated tools can enhance the identification of sexist content, they are often restricted to binary classification
- Consequently, more subtle manifestations of sexism may remain undetected due to the lack of fine-grained, context-sensitive labels
- To address this issue, we make the following contributions: (1) we present FineMuSe, a new multimodal sexism detection dataset in Spanish t…
- Our findings indicate that multimodal LLMs perform competitively with human annotators in identifying nuanced forms of sexism; however, the…
#arXiv #paper #研究/论文
14. A Note on Non-Composability of Layerwise Approximate Verification for Neural Inference
分类：研究/论文来源：arxiv_search分数：48作者：Or Zamir时间：2026-02-17T17:41:59Z
A natural and informal approach to verifiable (or zero-knowledge) ML inference over floating-point data is: ``prove that each layer was computed correctly up t…
- A natural and informal approach to verifiable (or zero-knowledge) ML inference over floating-point data is: ``prove that each layer was com…
- This short note gives a simple counterexample showing that this inference is false in general: for any neural network, we can construct a f…
#arXiv #paper #研究/论文
15. Beyond Match Maximization and Fairness: Retention-Optimized Two-Sided Matching
分类：研究/论文来源：arxiv_search分数：45作者：Ren Kishimoto时间：2026-02-17T17:30:53Z
On two-sided matching platforms such as online dating and recruiting, recommendation algorithms often aim to maximize the total number of matches. However, thi…
- On two-sided matching platforms such as online dating and recruiting, recommendation algorithms often aim to maximize the total number of m…
- However, this objective creates an imbalance, where some users receive far too many matches while many others receive very few and eventual…
- Retaining users is crucial for many platforms, such as those that depend heavily on subscriptions
- Some may use fairness objectives to solve the problem of match maximization
- However, fairness in itself is not the ultimate objective for many platforms, as users do not suddenly reward the platform simply because e…
- In practice, where user retention is often the ultimate goal, casually relying on fairness will leave the optimization of retention up to l…
#arXiv #paper #研究/论文
16. Enabling Low-Latency Machine learning on Radiation-Hard FPGAs with hls4ml
分类：研究/论文来源：arxiv_search分数：42作者：Katya Govorkova时间：2026-02-17T17:30:28Z
This paper presents the first demonstration of a viable, ultra-fast, radiation-hard machine learning (ML) application on FPGAs, which could be used in future h…
- This paper presents the first demonstration of a viable, ultra-fast, radiation-hard machine learning (ML) application on FPGAs, which could…
- We present a three-fold contribution, with the PicoCal calorimeter, planned for the LHCb Upgrade II experiment, used as a test case
- First, we develop a lightweight autoencoder to compress a 32-sample timing readout, representative of that of the PicoCal, into a two-dimen…
- Second, we introduce a systematic, hardware-aware quantization strategy and show that the model can be reduced to 10-bit weights with minim…
- Third, as a barrier to the adoption of on-detector ML is the lack of support for radiation-hard FPGAs in the High-Energy Physics community'…
- This new back-end enables the automatic translation of ML models into High-Level Synthesis (HLS) projects for the Microchip PolarFire famil…
#arXiv #paper #研究/论文
17. UrbanVerse: Learning Urban Region Representation Across Cities and Tasks
分类：研究/论文来源：arxiv_search分数：40作者：Fengze Sun时间：2026-02-17T17:28:48Z
Recent advances in urban region representation learning have enabled a wide range of applications in urban analytics, yet existing methods remain limited in th…
- Recent advances in urban region representation learning have enabled a wide range of applications in urban analytics, yet existing methods …
- We aim to generalize urban representation learning beyond city- and task-specific settings, towards a foundation-style model for urban anal…
- To this end, we propose UrbanVerse, a model for cross-city urban representation learning and cross-task urban analytics
- For cross-city generalization, UrbanVerse focuses on features local to the target regions and structural features of the nearby regions rat…
- We model regions as nodes on a graph, which enables a random walk-based procedure to form "sequences of regions" that reflect both local an…
- For cross-task generalization, we propose a cross-task learning module named HCondDiffCT
#arXiv #paper #研究/论文
18. MRC-GAT: A Meta-Relational Copula-Based Graph Attention Network for Interpretable Multimodal Alzheimer's Disease Diagnosis
分类：研究/论文来源：arxiv_search分数：38作者：Fatemeh Khalvandi时间：2026-02-17T17:15:32Z
Alzheimer's disease (AD) is a progressive neurodegenerative condition necessitating early and precise diagnosis to provide prompt clinical management. Given th…
- Alzheimer's disease (AD) is a progressive neurodegenerative condition necessitating early and precise diagnosis to provide prompt clinical …
- Given the paramount importance of early diagnosis, recent studies have increasingly focused on computer-aided diagnostic models to enhance …
- However, most graph-based approaches still rely on fixed structural designs, which restrict their flexibility and limit generalization acro…
- To overcome these limitations, the Meta-Relational Copula-Based Graph Attention Network (MRC-GAT) is proposed as an efficient multimodal mo…
- The proposed architecture, copula-based similarity alignment, relational attention, and node fusion are integrated as the core components o…
- According to evaluations performed on the TADPOLE and NACC datasets, the MRC-GAT model achieved accuracies of 96
#arXiv #paper #研究/论文
19. Beyond Labels: Information-Efficient Human-in-the-Loop Learning using Ranking and Selection Queries
分类：研究/论文来源：arxiv_search分数：35作者：Belén Martín-Urcelay时间：2026-02-17T17:14:15Z
Integrating human expertise into machine learning systems often reduces the role of experts to labeling oracles, a paradigm that limits the amount of informati…
- Integrating human expertise into machine learning systems often reduces the role of experts to labeling oracles, a paradigm that limits the…
- We address this challenge by developing a human-in-the-loop framework to learn binary classifiers with rich query types, consisting of item…
- We first introduce probabilistic human response models for these rich queries motivated by the relationship experimentally observed between…
- Using these models, we then design active learning algorithms that leverage the rich queries to increase the information gained per interac…
- We provide theoretical bounds on sample complexity and develop a tractable and computationally efficient variational approximation
- Through experiments with simulated annotators derived from crowdsourced word-sentiment and image-aesthetic datasets, we demonstrate signifi…
#arXiv #paper #研究/论文
20. MeshMimic: Geometry-Aware Humanoid Motion Learning through 3D Scene Reconstruction
分类：研究/论文来源：arxiv_search分数：32作者：Qiang Zhang时间：2026-02-17T17:09:45Z
Humanoid motion control has witnessed significant breakthroughs in recent years, with deep reinforcement learning (RL) emerging as a primary catalyst for achie…
- Humanoid motion control has witnessed significant breakthroughs in recent years, with deep reinforcement learning (RL) emerging as a primar…
- However, the high dimensionality and intricate dynamics of humanoid robots make manual motion design impractical, leading to a heavy relian…
- These datasets are not only costly to acquire but also frequently lack the necessary geometric context of the surrounding physical environm…
- Consequently, existing motion synthesis frameworks often suffer from a decoupling of motion and scene, resulting in physical inconsistencie…
- In this work, we present MeshMimic, an innovative framework that bridges 3D scene reconstruction and embodied intelligence to enable humano…
- By leveraging state-of-the-art 3D vision models, our framework precisely segments and reconstructs both human trajectories and the underlyi…
#arXiv #paper #研究/论文