Claude Fable 5.2 · Humanity's Last Exam(with tools;题集版本未注明) · 附件证据(ev-f52-hle-tools-scaling)
Claude Fable 5.2 的 Humanity's Last Exam · with tools;题集版本未注明 官方出处。见《Claude Fable 5.2 and Claude Mythos 5.2 System Card》PDF p.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签)。证据 ev-f52-hle-tools-scaling。
Claude Fable 5.2 · Humanity's Last Exam(with tools;题集版本未注明),HLE with tools / max配置,Reported score 74.2%(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.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签)。 查看成绩条件
Claude Fable 5.2 · Humanity's Last Exam(with tools;题集版本未注明),HLE with tools / low配置,Reported score 57.4%(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.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签)。 查看成绩条件
Claude Fable 5.2 · Humanity's Last Exam(with tools;题集版本未注明),HLE with tools / medium配置,Reported score 68.6%(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.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签)。 查看成绩条件
Claude Fable 5.2 · Humanity's Last Exam(with tools;题集版本未注明),HLE with tools / high配置,Reported score 71.5%(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.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签)。 查看成绩条件
Claude Fable 5.2 · Humanity's Last Exam(with tools;题集版本未注明),HLE with tools / xhigh配置,Reported score 73.4%(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.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签)。 查看成绩条件
附件 · PDF p.217, Figure 8.11.1.A, HLE with tools(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 · Humanity's Last Exam(with tools;题集版本未注明),HLE with tools / max配置,Reported score 74.2%(具体成绩公布时间未确认);Claude Fable 5.2 · Humanity's Last Exam(with tools;题集版本未注明),HLE with tools / low配置,Reported score 57.4%(具体成绩公布时间未确认);Claude Fable 5.2 · Humanity's Last Exam(with tools;题集版本未注明),HLE with tools / medium配置,Reported score 68.6%(具体成绩公布时间未确认);Claude Fable 5.2 · Humanity's Last Exam(with tools;题集版本未注明),HLE with tools / high配置,Reported score 71.5%(具体成绩公布时间未确认);Claude Fable 5.2 · Humanity's Last Exam(with tools;题集版本未注明),HLE with tools / xhigh配置,Reported score 73.4%(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.217, Figure 8.11.1.A, HLE with tools(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=217。Benchlog https://benchlog.lekshome.cc/evidence/claude-fable-and-mythos-5-2/ev-f52-hle-tools-scaling/
关联成绩与条件
Claude Fable 5.2 · Humanity's Last Exam 成绩与配置
指标与成绩
配置与评测条件
证据关系
Reported score74.2%原文 74.2% · 单位 %
公布 未确认
HLE with tools / max
设置
Claude Fable 5.2;thinking auto;max effort;web search, web fetch, programmatic tool calling, code execution;980k-token task budget;no context compaction。
未知项
该 effort 点的具体 trial 数未在图注列明。
工具
允许工具
推理预算
max
框架
未记录
协议
Humanity's Last Exam with tools;版本/工具配置未在主文表格列出。
条件
Thinking auto; web search/fetch, programmatic tools and code execution; 980k-token budget; no context compaction; Opus 4.6 grader; max effort.
Claude Fable 5.2 · Humanity's Last Exam (with tools;题集版本未注明) · Attachment evidence (ev-f52-hle-tools-scaling)
Official source for Humanity's Last Exam · with tools;题集版本未注明 in Claude Fable 5.2. See “Claude Fable 5.2 and Claude Mythos 5.2 System Card” at PDF p.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签). Evidence ev-f52-hle-tools-scaling.
Claude Fable 5.2 · Humanity's Last Exam (with tools;题集版本未注明), configuration HLE with tools / max, Reported score 74.2% (exact score publication date unconfirmed). See “Claude Fable 5.2 and Claude Mythos 5.2 System Card” at PDF p.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签). View score conditions
Claude Fable 5.2 · Humanity's Last Exam (with tools;题集版本未注明), configuration HLE with tools / low, Reported score 57.4% (exact score publication date unconfirmed). See “Claude Fable 5.2 and Claude Mythos 5.2 System Card” at PDF p.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签). View score conditions
Claude Fable 5.2 · Humanity's Last Exam (with tools;题集版本未注明), configuration HLE with tools / medium, Reported score 68.6% (exact score publication date unconfirmed). See “Claude Fable 5.2 and Claude Mythos 5.2 System Card” at PDF p.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签). View score conditions
Claude Fable 5.2 · Humanity's Last Exam (with tools;题集版本未注明), configuration HLE with tools / high, Reported score 71.5% (exact score publication date unconfirmed). See “Claude Fable 5.2 and Claude Mythos 5.2 System Card” at PDF p.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签). View score conditions
Claude Fable 5.2 · Humanity's Last Exam (with tools;题集版本未注明), configuration HLE with tools / xhigh, Reported score 73.4% (exact score publication date unconfirmed). See “Claude Fable 5.2 and Claude Mythos 5.2 System Card” at PDF p.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签). View score conditions
Attachment · PDF p.217, Figure 8.11.1.A, HLE with tools(已渲染并视觉核读数值标签)
Material version
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.
Version published
Disclosure date
Unconfirmed · Launch or later disclosure unconfirmed
Collected
Historical basis
No reliable historical version evidence available
Citation text
Claude Fable 5.2 · Humanity's Last Exam (with tools;题集版本未注明), configuration HLE with tools / max, Reported score 74.2% (exact score publication date unconfirmed); Claude Fable 5.2 · Humanity's Last Exam (with tools;题集版本未注明), configuration HLE with tools / low, Reported score 57.4% (exact score publication date unconfirmed); Claude Fable 5.2 · Humanity's Last Exam (with tools;题集版本未注明), configuration HLE with tools / medium, Reported score 68.6% (exact score publication date unconfirmed); Claude Fable 5.2 · Humanity's Last Exam (with tools;题集版本未注明), configuration HLE with tools / high, Reported score 71.5% (exact score publication date unconfirmed); Claude Fable 5.2 · Humanity's Last Exam (with tools;题集版本未注明), configuration HLE with tools / xhigh, Reported score 73.4% (exact score publication date unconfirmed). Official source: Claude Fable 5.2 and Claude Mythos 5.2 System Card (Attachment, 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.217, Figure 8.11.1.A, HLE with tools(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=217. Benchlog https://benchlog.lekshome.cc/evidence/claude-fable-and-mythos-5-2/ev-f52-hle-tools-scaling/?lang=en
Related scores and conditions
Claude Fable 5.2 · Humanity's Last Exam scores and configurations
Metric and score
Configuration and evaluation conditions
Evidence relation
Reported score74.2%Reported 74.2% · Unit %
Published Unconfirmed
HLE with tools / max
Settings
Claude Fable 5.2;thinking auto;max effort;web search, web fetch, programmatic tool calling, code execution;980k-token task budget;no context compaction。
Unknowns
该 effort 点的具体 trial 数未在图注列明。
Tools
Tools allowed
Reasoning budget
max
Framework
not recorded
Protocol
Humanity's Last Exam with tools;版本/工具配置未在主文表格列出。
Conditions
Thinking auto; web search/fetch, programmatic tools and code execution; 980k-token budget; no context compaction; Opus 4.6 grader; max effort.