Tag
RLRT is an on-device reinforcement learning runtime for embedded GPUs that makes actor-learner policy coupling a tunable parameter via a refresh bound and slack-aware GPU admission, keeping deadline misses below 1% versus 85-100% without it on Jetson boards.
Cal.com v7 launches with new features including events, discover, and an AI assistant alongside other improvements.
Postey.ai launches an AI-powered content management and scheduling tool that lets creators post to 9 platforms from Claude/ChatGPT, with built-in DM automation, a writing assistant, and cross-platform analytics.
LRMoE is a memory-aware MoE training system that jointly recalibrates backward demand and balances retained workload on memory-constrained GPU nodes, reducing cumulative MoE-layer training time by 21.2-44.8% on Qwen3-30B-A3B and GLM4.5-106B-A12B compared to SYMI.
Part 6 of an open-source Inference Engineering repo is released, covering KV cache memory allocation, growth, release, and capacity-aware scheduling using real model inference and DynamicCache.
TAO 提出一个工具感知的联合调度框架,用于基于 LLM 的智能体服务,通过动态规划在容量约束下选择和放置会话,相比最优基线最高降低 26.8% 延迟、36.3% TTFT 并提升 29.5% 解码吞吐。
A Lobsters discussion about using timers to perform weekly tasks, likely covering programming or system administration techniques for scheduling recurring work.
RAC 提出了一种与执行框架无关的“关系型智能体合约”,让智能体只声明语义关系、由专用工具推导具体实现,并在 PCB 设计、作业车间调度和嵌入式代码任务上显著提升了约束满足率(如 PCB 平均 90.9% 对比 74.0%)。
提出 VISTA 这一免训练方法,通过在少量全引导 warmup 生成中收集价值统计并用动态规划优化,在固定预算下自动为离散扩散的 SMC 奖励引导选择引导时间步,在文本、生物序列和图像生成任务上优于手工稀疏调度,实现最高 5.35× 加速。
MeetPaw is a lightweight macOS menu-bar app that generates instant Google Meet, Zoom, or Microsoft Teams links with one click, copying them to the clipboard for frictionless sharing.
Manus 宣布支持将重复性工作设为自动执行(autopilot),用户只需指定工作启动时间和期望的完成结果,即可实现任务自动化。
Calenteri 是一款在 Product Hunt 上发布的 SaaS 预约工具,主打 GDPR 合规、欧盟托管且无第三方追踪器,作为 Calendly 的隐私友好替代方案,支持用户间互订日程并邮件通知。
CostGraph 博客详解如何把 Kubernetes 节点资源整理建模为多维向量装箱问题(d 维 vector bin packing),说明为何按单一尺寸排序(first fit decreasing)会浪费容量,并介绍基于 Panigrahy et al. (2011) 几何启发式、结合真实用量 P95 与调度器硬性约束的贪心打包方案。
一项受控等预算实验研究比较了蒙特卡洛树搜索(MCTS)与随机重启列表调度在DAG工作流调度中的表现,发现MCTS在相同预算下反而比简单的随机重启方法差约2.88%。研究指出MCTS的劣势源于选择规则锁定在首个扩展子节点上,而换成标准UCT规则后差距基本消失。
BrightBean Studio 是一款免费开源的社交媒体管理工具,作为 Buffer 的替代方案,支持向 10+ 平台排程发布,并内置 MCP 服务器和 REST API 以接入 AI 智能体。
The paper introduces NP-Bench, an environment-grounded benchmark for verifying coordination among parallel LLM coding agents, and a proactive scheduling planner built into Nerveplane that partitions work into disjoint scopes and orders merges along dependency graphs. The planner lifts clean-integration success from 1/9 to 9/9 scenarios, eliminates merge conflicts, and collapses repeated-mistake rates from 1.00 to 0.00 via cross-session memory.
TokenRouter 是一个面向 token 级 LLM 路由的高效推理服务系统,采用「请求中心编程、模型中心执行」架构与延迟批调度策略,实现较现有系统 2.01–64.15 倍的解码吞吐提升。
PostNitro 推出「PostNitro for Agents」,让 AI 代理(如 Claude、ChatGPT)通过 OAuth 一键连接、35 工具 MCP 服务器、CLI 或 REST API 自动生成并排期发布品牌化的轮播图、图片和短视频到各大社交平台。
Meta is open-sourcing Rebalancer, an assignment-problem solver used internally for nine years to handle resource allocation across its infrastructure, including hardware placement, service/task placement, and traffic routing. It offers both an optimal MIP solver and a parallelized local search solver, handling ~40 million problems daily at Meta.
An educational walkthrough of the LLM serving scheduling loop, explaining how inference requests move through waiting queues, batch construction, prefill, decode, and continuous batching to keep GPUs saturated as requests finish at different times.