Throughput vs. Goodput: The Performance Metricin LLM Testing
用卖Dosa的摊位类比,讲清LLM测试中常被忽略的吞吐量与有效吞吐量区别
Article URL: https://qainsights.com/throughput-vs-goodput-the-performance-metric-you-are-probably-ignoring-in-llm-testing/ Comments URL: https://news.…
用卖Dosa的摊位类比,讲清LLM测试中常被忽略的吞吐量与有效吞吐量区别
Article URL: https://qainsights.com/throughput-vs-goodput-the-performance-metric-you-are-probably-ignoring-in-llm-testing/ Comments URL: https://news.…
面向Superchip的LLM推理优化,提出SLO感知的旋转调度与内存管理方案,已被MLSys '26接收。
arXiv:2601.20309v2 Announce Type: replace-cross Abstract: Large Language Model (LLM) serving faces a fundamental tension between stringent latency Ser…
面向共享GPU集群,提出连续自适应方法优化大模型服务SLO,降低延迟与成本
arXiv:2604.16400v2 Announce Type: replace-cross Abstract: As Large Language Models (LLMs) are increasingly adopted in edge intelligence to power domai…
面向大规模复合AI服务的SLO感知查询规划器,提出保障响应时间与服务质量的新方案。
arXiv:2504.16397v2 Announce Type: replace-cross Abstract: The rise of compound AI serving that integrates multiple operators in a pipeline enables end…
最新研究揭示2026年大模型代码生成中包幻觉仍存,最高21.7%的虚假包名成slopsquatting攻击新入口
arXiv:2605.17062v1 Announce Type: cross Abstract: Spracklen et al. (USENIX Security '25) showed that code-generating large language models hallucinate…
Robin Sloan解读M.F.K. Fisher的写作魔力,用“声音”唤醒小说的灵魂,文学爱好者的灵感之泉。
Robin Sloan on his novel Sourdough and the author M.F.K Fischer for FSG : And for me, voice is the thing. In a novel, I will forgive any flaw, overloo…
探讨互联网中AI生成的低质“垃圾内容”泛滥程度,以及它如何催化“脑腐”文化,拷问数字生态的未来。
Article URL: https://www.statsignificant.com/p/how-much-of-the-internet-is-ai-slop Comments URL: https://news.ycombinator.com/item?id=48207413 Points:…
一个确定性质量门控工具,专门验证AI生成的代码,无需依赖大模型,MIT开源。
Article URL: https://github.com/scanaislop/aislop Comments URL: https://news.ycombinator.com/item?id=48191872 Points: 2 # Comments: 0
面向LLM推理的SLO感知旋转调度与内存管理技术,已被MLSys 2026录用,值得关注。
Article URL: https://supercomputing-system-ai-lab.github.io/projects/superinfer/ Comments URL: https://news.ycombinator.com/item?id=48188146 Points: 3…
SvelteKit公测版发布,slotted components功能上线,前端开发者不容错过的框架更新。
Two projects that have been months (even years) in the making have made their way out into the world. SvelteKit is now in public beta and slotted comp…
专为AI slop构建的公开数据集,帮AI学会避开低质量输出
I just had this idea, you read it all the time AI slop is so prevalent people are getting banned for a year for submitting science papers to arXiv wit…
新型OLAP数据库SlothDB在Clickbench基准测试中超越DuckDB,基于C++20实现且完全开源。
Beats Clickbench 43 100M Parquet on 33 Queries. Code - https://github.com/SouravRoy-ETL/slothdb Comments URL: https://news.ycombinator.com/item?id=481…
提出Slot-MPC,用物体中心表示替代传统特征,将世界模型与梯度模型预测控制结合,使机器人能在未见场景中高效规划动作,性能与效率双赢,是目标条件控制的新范式。
arXiv:2605.14937v1 Announce Type: cross Abstract: Predictive world models enable agents to model scene dynamics and reason about the consequences of t…