Harvey LAB-AA · Artificial Analysis v1.1; 120-task held-out set; hallucination-gated and ungated all-pass rates · 测试定义与成绩

Harvey LAB-AA · Artificial Analysis v1.1; 120-task held-out set; hallucination-gated and ungated all-pass rates 收录 3 次模型发布的评测披露或成绩。

测试版本与指标

别名 Harvey LAB-AA v1.1

  • Hallucination-gated all-pass rate · % · 越高越好
  • All-pass rate before hallucination gate · % · 越高越好

各系列最新模型 · 公布成绩

常规展示范围。排名按指标与评测协议分组,已记录的工具、预算及框架分别展示。条件未说明的组保留公布值供查证。

Hallucination-gated all-pass rate %

工具条件未记录 / low / 框架未记录 · 条件未说明

Harvey LAB-AA · Artificial Analysis v1.1; 120-task held-out set; hallucination-gated and ungated all-pass rates · Hallucination-gated all-pass rate · 工具条件未记录 / low / 框架未记录 · 条件未说明
模型与配置成绩(%)记录与证据
GPT-6.2 SolLow3.1%
GPT-6.2 AstraLow2.2%

工具条件未记录 / medium / 框架未记录 · 条件未说明

Harvey LAB-AA · Artificial Analysis v1.1; 120-task held-out set; hallucination-gated and ungated all-pass rates · Hallucination-gated all-pass rate · 工具条件未记录 / medium / 框架未记录 · 条件未说明
模型与配置成绩(%)记录与证据
GPT-6.2 AstraMedium8.1%
GPT-6.2 SolMedium5%

工具条件未记录 / high / 框架未记录 · 条件未说明

Harvey LAB-AA · Artificial Analysis v1.1; 120-task held-out set; hallucination-gated and ungated all-pass rates · Hallucination-gated all-pass rate · 工具条件未记录 / high / 框架未记录 · 条件未说明
模型与配置成绩(%)记录与证据
GPT-6.2 AstraHigh13.3%
GPT-6.2 SolHigh5.8%

工具条件未记录 / xhigh / 框架未记录 · 条件未说明

Harvey LAB-AA · Artificial Analysis v1.1; 120-task held-out set; hallucination-gated and ungated all-pass rates · Hallucination-gated all-pass rate · 工具条件未记录 / xhigh / 框架未记录 · 条件未说明
模型与配置成绩(%)记录与证据
GPT-6.2 AstraXhigh12.8%
GPT-6.2 SolXhigh6.1%

工具条件未记录 / max / 框架未记录 · 条件未说明

Harvey LAB-AA · Artificial Analysis v1.1; 120-task held-out set; hallucination-gated and ungated all-pass rates · Hallucination-gated all-pass rate · 工具条件未记录 / max / 框架未记录 · 条件未说明
模型与配置成绩(%)记录与证据
GPT-6.2 AstraMax15%
GPT-6.2 SolMax11.1%

工具条件未记录 / max / 框架未记录 · Harvey LAB-AA v1.1 held-out 120-task set; Artificial Analysis evaluation; max effort; average across three judge models.

Harvey LAB-AA · Artificial Analysis v1.1; 120-task held-out set; hallucination-gated and ungated all-pass rates · Hallucination-gated all-pass rate · 工具条件未记录 / max / 框架未记录 · Harvey LAB-AA v1.1 held-out 120-task set; Artificial Analysis evaluation; max effort; average across three judge models.
模型与配置成绩(%)记录与证据
Claude Fable 5.2Max7.2%

All-pass rate before hallucination gate %

工具条件未记录 / max / 框架未记录 · Harvey LAB-AA v1.1 held-out 120-task set; Artificial Analysis evaluation; max effort; average across three judge models.

Harvey LAB-AA · Artificial Analysis v1.1; 120-task held-out set; hallucination-gated and ungated all-pass rates · All-pass rate before hallucination gate · 工具条件未记录 / max / 框架未记录 · Harvey LAB-AA v1.1 held-out 120-task set; Artificial Analysis evaluation; max effort; average across three judge models.
模型与配置成绩(%)记录与证据
Claude Fable 5.2Max10.8%

历史记录 · 3 次发布

  • GPT-6.2 Astra本次发布有 5 条成绩记录,含 5 条精确值(可含历史记录)
  • GPT-6.2 Sol本次发布有 5 条成绩记录,含 5 条精确值(可含历史记录)
  • Claude Fable 5.2本次发布有 2 条成绩记录,含 2 条精确值(可含历史记录)