Claude Opus 4.5 · Simple-QA-Verified · Anthropic factuality evaluation; Claude Sonnet 4.5 grader; exact dataset revision not stated · 成绩与配置

Claude Opus 4.5 的 Simple-QA-Verified · Anthropic factuality evaluation; Claude Sonnet 4.5 grader; exact dataset revision not stated 记录。本次发布有 6 条成绩记录,含 6 条精确值(可含历史记录),附 1 条来源定位。

成绩与评测条件

Claude Opus 4.5 · Simple-QA-Verified 成绩与配置
指标与成绩配置与评测条件出处
Correct rate33.1%原文 33.1% · 单位 %
公布 未确认
Public factuality datasets, no thinking
设置
No thinking; no external tools stated; Claude Sonnet 4.5 grades Simple-QA-Verified and AA-Omniscience.
未知项
Trial count and model sampling settings are not stated.
工具
工具条件未记录
推理预算
未记录
框架
未记录
协议
Anthropic Simple-QA-Verified factuality evaluation; correct rate; Sonnet 4.5 grader.
条件
No thinking; trial count and sampling settings not stated.
在排名中查看此成绩
Incorrect rate36.4%原文 36.4% · 单位 %
公布 未确认
Public factuality datasets, no thinking
设置
No thinking; no external tools stated; Claude Sonnet 4.5 grades Simple-QA-Verified and AA-Omniscience.
未知项
Trial count and model sampling settings are not stated.
工具
工具条件未记录
推理预算
未记录
框架
未记录
协议
Anthropic Simple-QA-Verified factuality evaluation; incorrect rate; Sonnet 4.5 grader.
条件
No thinking; trial count and sampling settings not stated.
在排名中查看此成绩
Uncertain response rate30.6%原文 30.6% · 单位 %
公布 未确认
Public factuality datasets, no thinking
设置
No thinking; no external tools stated; Claude Sonnet 4.5 grades Simple-QA-Verified and AA-Omniscience.
未知项
Trial count and model sampling settings are not stated.
工具
工具条件未记录
推理预算
未记录
框架
未记录
协议
Anthropic Simple-QA-Verified factuality evaluation; uncertain response rate; Sonnet 4.5 grader.
条件
No thinking; trial count and sampling settings not stated.
在排名中查看此成绩
Correct rate44.1%原文 44.1% · 单位 %
公布 未确认
Public factuality datasets, 16K extended thinking
设置
16K extended-thinking budget; no external tools stated; Claude Sonnet 4.5 grades Simple-QA-Verified and AA-Omniscience.
未知项
Trial count and model sampling settings are not stated.
工具
工具条件未记录
推理预算
思考预算 16K
框架
未记录
协议
Anthropic Simple-QA-Verified factuality evaluation; correct rate; Sonnet 4.5 grader.
条件
16K extended-thinking budget; trial count and sampling settings not stated.
在排名中查看此成绩
Incorrect rate35.4%原文 35.4% · 单位 %
公布 未确认
Public factuality datasets, 16K extended thinking
设置
16K extended-thinking budget; no external tools stated; Claude Sonnet 4.5 grades Simple-QA-Verified and AA-Omniscience.
未知项
Trial count and model sampling settings are not stated.
工具
工具条件未记录
推理预算
思考预算 16K
框架
未记录
协议
Anthropic Simple-QA-Verified factuality evaluation; incorrect rate; Sonnet 4.5 grader.
条件
16K extended-thinking budget; trial count and sampling settings not stated.
在排名中查看此成绩
Uncertain response rate20.5%原文 20.5% · 单位 %
公布 未确认
Public factuality datasets, 16K extended thinking
设置
16K extended-thinking budget; no external tools stated; Claude Sonnet 4.5 grades Simple-QA-Verified and AA-Omniscience.
未知项
Trial count and model sampling settings are not stated.
工具
工具条件未记录
推理预算
思考预算 16K
框架
未记录
协议
Anthropic Simple-QA-Verified factuality evaluation; uncertain response rate; Sonnet 4.5 grader.
条件
16K extended-thinking budget; trial count and sampling settings not stated.
在排名中查看此成绩

出处与披露

测试版本与指标

  • Correct rate · % · 越高越好
  • Incorrect rate · % · 越低越好
  • Uncertain response rate · % · 越低越好