2026年8月15日 技术热点总结

项目简介：Plugin and skin collection for DeepSeek Harness (DSH) Web UI – task board, git graph, right-side panel, remote mobile UI, pet, live token stats, and skin center.

技术栈：TypeScript、deepseek-harness、dsh、dsh-plugin、web-ui

核心介绍：中文 | English dsh-web-ui 是 DeepSeek Harness（DSH）Web UI 的插件与皮肤集合：任务看板、Git 图谱、右侧面板、移动端远程、远程连接、图像理解工具、鲸鱼娘宠物、实时令牌统计，以及皮肤中心。所有插件既可独立安装，也可通过聚合包一次装齐。 在侧边栏点击「任务看板」进入。任务按五列状态组织：待规划、待办、进行中、已完成、已失败。点击卡片上的「执行」，任务将由真实的 DSH 智能体会话执行，完成后状态自动回写；需要复盘时，可直接跳转到执行会话查看完整过程。 任务支持定时执行：在详情中配置 cron 表达式（如每天 23:00 自动升级 DSH、每周一 09:00 生成周报），到点自动开工，无需人工值守。 输入框上方的分支选择器，支持切换分支与查看提交历史；Git 图…

项目简介：为 DeepSeek Harness (DSH) 生态打造的现代化桌面端体验

技术栈：TypeScript、deepseek、deepseek-harness、desktop、dsh、dsh-plugin、dsh-plugin-desktop

核心介绍：DeepSeek Harness Desktop 中文 · English 为DeepSeek Harness生态打造的现代化桌面端体验（插件） DeepSeek Harness 官方目前通过命令行启动本地 Web UI。本项目将服务启动、运行管理和桌面窗口整合为开箱即用的桌面体验，让用户无需配置 Node.js 或执行命令，即可直接使用。 Download Desktop 把官方 DeepSeek Harness 的本地 Web UI 带到原生桌面。应用自动启动和管理本地 Harness 服务，集成系统托盘与桌面窗口，无需安装 Node.js 或执行命令。 手机远程控制 通过 iOS 和 Android 远程连接 Desktop，在手机上发起任务、查看 Agent 进度，并在需要时继续跟进。 插件市场 Harness 遵循“一切皆插件”的架构。桌面端插件市场将提供插件的发现、安装、更新

项目数据：⭐ 1,817 Stars，🍴 94 Forks

项目简介：A Programming Paradigm for Spatiotemporal Composability

核心介绍：> This is a preprint under active revision. The content may change substantially; please cite the latest version and check back before relying on specific results.

项目数据：⭐ 1,311 Stars，🍴 46 Forks

项目简介：A curated list of plugins for DeepSeek Harness (dsh) · DeepSeek Harness 插件精选列表

技术栈：Python、awesome、awesome-list、deepseek-harness、dsh、dsh-plugin

核心介绍：English | 中文 > A curated list of plugins for DeepSeek Harness (dsh). DeepSeek Harness is DeepSeek’s open-source agent harness — a runnable coding agent (Web and headless), built on a framework where everything is a plugin: models, tools, sandboxes, session storage, UI, even the agent loop itself. Plugins can extend the official coding agent, swap out its core parts, or assemble something entirely different.

项目数据：⭐ 1,092 S…

项目简介：解决DSH 官方尚无终端 TUI 痛点的补位之作，献给偏爱cli的各位极客：Claude Code 风格全屏交互终端插件——像素鲸鱼顶栏、实时工作状态行、思考流式展开、双击 Esc 回滚、上下文进度条 + TPS 仪表。npm 一键安装。

技术栈：TypeScript、claude-code、coding-agent、deepseek、deepseek-harness、dsh-plugin、ink、react、terminal

核心介绍：简体中文 | English > 一个美观且实用的 Claude Code 风格 TUI 插件：像素鲸鱼顶栏、双流光大字、实时工作状态行、思考流式展开、双击 Esc 时间回溯、蓝白上下文进度条 + TPS 仪表。 > 零核心改动，纯插件挂载。安装插件即可启用，卸载后不会留下核心补丁。 本插件被 DeepSeek Harness 官方公众号 推文收录，作为”内测用户精选插件”展示： （消息任意位置补全，发送时自动附加文件内容/目录列表）、历史搜索、消息选择、 inline/alternate-screen 两种渲染模式，以及 /lang 中英界面语言切换。 推理等级、输入/输出 token 与 Git/会话信息。 以及双击 Esc 发起的会话 rewind/fork。 ask_user_…

We introduce , a recurrent Transformer architecture with fixed-size memory that generalizes sliding-window attention while remaining parallelizable during training. consists of two coupled models: a prefiller Q, which leverages full attentionIn practice, we use interleaved full and sliding-window attention for Q, as this yields stronger performance. The essential requirement is that Q be more expressive than P, with access to the full history. to produce memory targets m'_t, and a decoder P, which uses only sliding-window attention and recurrent K/V injection to produce decoder memories m_…

Test-time compute scaling is a primary driver of performance in large reasoning models (LRMs), but extreme inefficiency bounds current approaches, shifting the critical question from how much compute to spend, to where to allocate it. We formalize test-time reasoning as a constrained compute allocation problem over partial trajectories. Under a fixed hardware budget, existing paradigms fail to actively allocate the compute to the most promising partial progress: traditional parallel sampling treats traces independently and induces severe memory bottlenecks, while subtractive pruning starves…

Rib fractures are common and time-consuming to localize on computed tomography (CT). We ask whether fractures detected independently in two orthogonal CT-derived projections (anteroposterior and lateral) can be paired across views and triangulated into reliable 3D points at a controlled rate of false outputs, and we answer it with a staged diagnostic study. The projection geometry is exact, and given correct correspondence, localization is accurate (median 4.0 mm, 88% within 10 mm, 93.6% rib-exact). On a sealed 55-case cohort, a large share of fractures is in principle recoverable (61.1% du…

Interactive autoregressive video generation demands both low-latency rollouts and precise online control. Few-step distillation accelerates generation by reducing denoising steps, while online control imposes a causal constraint: frames and blocks should depend on history and controls available during generation. Existing video distribution matching distillation (DMD) pipelines, however, often supervise causal few-step students using bidirectional teachers that score complete clips. The score for a target can therefore depend on future frames and controls that were unavailable when the stud…

Large audio-language models (LALMs) have demonstrated strong capabilities in understanding diverse audio inputs. This diversity includes low-frequency signals that are inaudible to humans but can still enter the model and influence its generation. However, the practical impact of such low-frequency inputs on LALMs remains largely unexplored. In this paper, we propose Intermittent Low-Frequency Lockout (ILL), an inaudible red teaming method that evaluates this risk using a universal waveform template in a black box setting. ILL uses Sentence Attention Scale Estimation to determine active int…

Machine translation (MT) systems often fail to correctly translate gender, especially when converting from a gender-neutral language like English to a gendered target language such as Romanian. This bias results in translations that default to masculine forms or reinforce gender stereotypes. We propose a hybrid pipeline to mitigate this issue by combining large language model (LLM)-based gender classification with neural machine translation (NMT). Our system uses a fine-tuned LLM to detect the intended gender of target words in English sentences and insert inline gender hint tags. These tag…