1
最早可追溯到 Win 3.1，微软资深工程师详解批处理文件 WINSTART.BAT
IT之家 8 月 12 日消息，据外媒 neowin 今天（12 日）报道，微软在长期改进 Windows 的过程中，也为操作系统积累了不少鲜为人知的功能。微软资深工程师雷蒙德 · 陈近日介绍了其中一个 大多数现代用户可能从未听说 过的功能，就是批处理文件 WINSTART.BAT 。 雷蒙德 · …
2
Request-Level Energy Attribution for Batched LLM Serving
批处理LLM服务能耗如何精确到每个请求？这篇论文提出请求级能量归因方法，为绿色AI提供量化新工具。
arXiv:2608.00026v1 Announce Type: new Abstract: Batched LLM serving improves throughput but complicates energy accounting. GPU power telemetry is aggr…
3
BlockServe: Block-Grained Continuous Batching for High-Throughput Diffusion LLM Serving
扩散LLM推理新方法：块级连续批处理突破吞吐瓶颈，论文提出BlockServe方案优化服务效率。
arXiv:2607.08930v1 Announce Type: new Abstract: Efficient serving of diffusion large language models (dLLMs) is hindered by convergence heterogeneity:…
4
How I Built a Bulk Video Processing Pipeline with Next.js, BullMQ, and FFmpeg
用Next.js+BullMQ+FFmpeg搭建并行视频处理管道，50个视频同时去静音、加字幕、裁剪，服务器不崩的实战方案
I've been building BatchEdits — a tool that lets content creators upload 50 videos at once and get them back edited simultaneously with silence remove…
5
Online Dynamic Batching with Formal Guarantees for LLM Training
形式化保证了LLM训练中在线动态批处理的高效性与稳定性，理论扎实实验详尽。
arXiv:2606.19989v1 Announce Type: cross Abstract: Modern LLM training breaks a core assumption behind offline batch samplers: the true training cost o…
6
Show HN: Durable, asynchronous LLM workflow engine in Rust
用 Rust 构建的持久化异步 LLM 工作流引擎，支持在线/批处理无缝切换，专为简化开发而生。
I built this over the weekend to scratch a personal itch. Might be useful to others. Supports transparently switching live to batch without code chang…
7
LLM Serving and the Bus That Never Stops
深入剖析LLM服务中永不停止的“飞行中批处理”机制，解锁推理效率提升新思路
Article URL: https://joker666.github.io/blog/2026-06-02-llm-serving-in-flight-batching Comments URL: https://news.ycombinator.com/item?id=48414773 Poi…
8
Video-Rate Streaming Stylization on a Vision-Aware MLLM-Conditioned Edit Diffusion: Asymmetric Batched Inference on a Distilled UNet + MLLM Text Encoder
视频流实时风格化新方法，结合视觉感知MLLM与扩散模型，实现非对称批处理推理，性能提升显著。
arXiv:2606.05981v1 Announce Type: new Abstract: Aggressive distillation of the diffusion U-Net inverts the per-frame bottleneck of real-time text-to-i…
9
Threshold-Based Exclusive Batching for LLM Inference
挑战LLM推理中的混合批处理惯例，全新阈值独占批处理策略或改写效率规则。
arXiv:2606.00516v1 Announce Type: new Abstract: Mixed batching (MB)--interleaving prefill and decode in a single batch--has become the standard schedu…
10
Memory-Bound but Not Bandwidth-Limited: The Physical AI Inference Gap in Batch-1 LLM Decode
揭示LLM推理瓶颈新视角：batch-1解码受内存限制而非带宽限制，挑战传统认知。
arXiv:2605.30571v1 Announce Type: cross Abstract: Physical AI systems, including robots, autonomous vehicles, embodied agents and edge copilots, often…
11
A Paired Testing Protocol for Batch-Conditioned Refusal Robustness in LLM Serving
提出批处理条件下LLM拒绝鲁棒性的配对测试协议，为安全评估提供更可靠的统计方法
arXiv:2605.27763v1 Announce Type: new Abstract: Safety evaluations of language models often treat serving configuration as fixed background infrastruc…
12
Show HN: IgniteMS – batch text embeddings at 253K msg/s on 8x A100
将文本嵌入批处理速度提升至25万条/秒，比Hugging Face TEI快3倍，开源且支持生产环境调优。
Article URL: https://github.com/Artain-AI/ignite-ms Comments URL: https://news.ycombinator.com/item?id=48210818 Points: 2 # Comments: 0