跳到主要内容Skip to main content

总分 / 综合评测Aggregate Score

Aggregate Score

  • 选型与成本Selection & cost
  • 第六站 · 选得对Station 6 · Choosing Right
解决什么问题What problem it solves

把多个基准的分数加权成一个总分,方便快速比较。Collapse several benchmark scores into one number for quick comparison.

这是什么What it is

多个评测维度的加权汇总。不同机构的加权方式差别很大。

A weighted aggregate across several evaluation dimensions. Weighting differs substantially between providers.

为什么需要知道Why it matters

看到「综合分 XX」要先问三件事:怎么加权?用的哪套题?什么时候的数据?When you see "an overall score of X", ask first: how was it weighted, which question set, and when was the data collected.

出处Sources
  • HELM 论文Holistic Evaluation of Language Models (HELM) ↗
  • Project MPG:多基准聚合Project MPG: generalized LLM benchmark ↗