AI产品精选

Weaviate Search Mode新增effort参数,控制测试时计算成本

Scaling Test-Time Compute is the easiest way to get higher quality outputs from AI systems, but it c...

精选理由

Weaviate给搜索模式加了effort档位,想省算力就选medium,要质量就上ultrahigh,BRIGHT基准上提升明显,做RAG的可以试试。

AI 摘要

Weaviate在Search Mode中引入effort参数,允许用户在medium、high、ultrahigh三档间选择,以权衡输出质量与计算成本。ultrahigh档在BRIGHT基准的5个子集上平均nDCG@10达到X,比medium提升Y,比混合搜索提升Z。该功能旨在让开发者根据应用需求灵活调整推理资源。

原文 · Weaviate

Scaling Test-Time Compute is the easiest way to get higher quality outputs from AI systems, but it c...

Scaling Test-Time Compute is the easiest way to get higher quality outputs from AI systems, but it comes at a cost! ⚖️ How much extra compute you want to trade for quality depends on your application. Introducing the effort parameter in Search Mode, giving you the power to control this tradeoff! Choose between “medium”, “high”, and “ultrahigh”. 🚀 Search Mode with ultrahigh effort achieves X nDCG @10 averaged across 5 subsets of the BRIGHT benchmark, improving by Y over medium effort and Z over hybrid search alone. Learn more in our latest blog post: weaviate.io/blog/search-mo… F 💬 1 🔄 3 ❤️ 6 👀 209 📊 4 ⚡

Weaviate Search Mode新增effort参数,控制测试时计算成本 · AI 热点