Claude Fable 5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations) · 附件证据(ev-f52-healthbench)

Claude Fable 5.2 的 HealthBench · 48,000+ rubric items across 5,000 multi-turn conversations 官方出处。见《Claude Fable 5.2 and Claude Mythos 5.2 System Card》PDF p.245, Figure 8.16.1.A, HealthBench(已渲染并视觉核读数值标签)。证据 ev-f52-healthbench。

原文与定位

Claude Fable 5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations),Adaptive Max配置,Raw score 58.8%(具体成绩公布时间未确认)。见《Claude Fable 5.2 and Claude Mythos 5.2 System Card》PDF p.245, Figure 8.16.1.A, HealthBench(已渲染并视觉核读数值标签)。 查看成绩条件

Claude Fable 5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations),Adaptive Max配置,Length-adjusted score 59.6%(具体成绩公布时间未确认)。见《Claude Fable 5.2 and Claude Mythos 5.2 System Card》PDF p.245, Figure 8.16.1.A, HealthBench(已渲染并视觉核读数值标签)。 查看成绩条件

Claude Fable 5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations),Low effort配置,Raw score 60.1%(具体成绩公布时间未确认)。见《Claude Fable 5.2 and Claude Mythos 5.2 System Card》PDF p.245, Figure 8.16.1.A, HealthBench(已渲染并视觉核读数值标签)。 查看成绩条件

Claude Fable 5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations),Low effort配置,Length-adjusted score 60.4%(具体成绩公布时间未确认)。见《Claude Fable 5.2 and Claude Mythos 5.2 System Card》PDF p.245, Figure 8.16.1.A, HealthBench(已渲染并视觉核读数值标签)。 查看成绩条件

Artificial Analysis scores graded by GPT-4.1: at max, Fable 5.2 raw 58.8% and length-adjusted 59.6%; at low, raw 60.1% and adjusted 60.4%. Safety-classifier fallbacks were enabled.
官方材料
Claude Fable 5.2 and Claude Mythos 5.2 System Card · PDF 第 245 页
定位
附件 · PDF p.245, Figure 8.16.1.A, HealthBench(已渲染并视觉核读数值标签)
材料版本
Official System Card; 264-page PDF, cover dated November 9, 2026; current PDF/TXT saved as 本地留存副本 and .txt; PDF pp.1-264 text read page-by-page; all 80 image/vector-figure pages in pp.1-206 were visually checked in 21 sheets, and capability figures in pp.207-254 were visually checked; no unread pages or figures remain.
版本公布
披露时间
未确认 · 首发/后补未确认
采集时间
历史依据
暂无可靠历史版本依据
引用文本

Claude Fable 5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations),Adaptive Max配置,Raw score 58.8%(具体成绩公布时间未确认);Claude Fable 5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations),Adaptive Max配置,Length-adjusted score 59.6%(具体成绩公布时间未确认);Claude Fable 5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations),Low effort配置,Raw score 60.1%(具体成绩公布时间未确认);Claude Fable 5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations),Low effort配置,Length-adjusted score 60.4%(具体成绩公布时间未确认)。官方出处 Claude Fable 5.2 and Claude Mythos 5.2 System Card(附件,Official System Card; 264-page PDF, cover dated November 9, 2026; current PDF/TXT saved as 本地留存副本 and .txt; PDF pp.1-264 text read page-by-page; all 80 image/vector-figure pages in pp.1-206 were visually checked in 21 sheets, and capability figures in pp.207-254 were visually checked; no unread pages or figures remain.)PDF p.245, Figure 8.16.1.A, HealthBench(已渲染并视觉核读数值标签),https://www.anthropic.com/document/claude-fable-5-2-mythos-5-2-system-card#page=245。Benchlog https://benchlog.lekshome.cc/evidence/claude-fable-and-mythos-5-2/ev-f52-healthbench/

关联成绩与条件

Claude Fable 5.2 · HealthBench 成绩与配置
指标与成绩配置与评测条件证据关系
Raw score58.8%原文 58.8% · 单位 %
公布 未确认
Adaptive Max
设置
Claude Fable 5.2;adaptive thinking always on;Max effort;默认 temperature/top_p;Card 通用表说明通常五次运行平均,个别测试另行注明。
未知项
System Card pp.1-264 文字已完整核读;各benchmark章节未明确披露的运行参数仍未知。
工具
工具条件未记录
推理预算
max
框架
未记录
协议
HealthBench; Artificial Analysis; GPT-4.1 grader; max effort.
条件
Safety-classifier fallbacks enabled; scores are not comparable with earlier Anthropic system cards graded by Opus 4.8.

直接支持该成绩

在排名中查看此成绩
Length-adjusted score59.6%原文 59.6% · 单位 %
公布 未确认
Adaptive Max
设置
Claude Fable 5.2;adaptive thinking always on;Max effort;默认 temperature/top_p;Card 通用表说明通常五次运行平均,个别测试另行注明。
未知项
System Card pp.1-264 文字已完整核读;各benchmark章节未明确披露的运行参数仍未知。
工具
工具条件未记录
推理预算
max
框架
未记录
协议
HealthBench length-adjusted score; Artificial Analysis; GPT-4.1 grader; max effort.
条件
Safety-classifier fallbacks enabled; scores are not comparable with earlier Anthropic system cards graded by Opus 4.8.

直接支持该成绩

在排名中查看此成绩
Raw score60.1%原文 60.1% · 单位 %
公布 未确认
Low effort
设置
Claude Fable 5.2;adaptive thinking always on;Low effort。
未知项
未另记
工具
工具条件未记录
推理预算
low
框架
未记录
协议
HealthBench; Artificial Analysis; GPT-4.1 grader; low effort.
条件
Safety-classifier fallbacks enabled; scores are not comparable with earlier Anthropic system cards graded by Opus 4.8.

直接支持该成绩

在排名中查看此成绩
Length-adjusted score60.4%原文 60.4% · 单位 %
公布 未确认
Low effort
设置
Claude Fable 5.2;adaptive thinking always on;Low effort。
未知项
未另记
工具
工具条件未记录
推理预算
low
框架
未记录
协议
HealthBench length-adjusted score; Artificial Analysis; GPT-4.1 grader; low effort.
条件
Safety-classifier fallbacks enabled; scores are not comparable with earlier Anthropic system cards graded by Opus 4.8.

直接支持该成绩

在排名中查看此成绩