Claude Sonnet 4.6 · BrowseComp(1,266 题;单代理与多代理按协议区分) · 附件证据(ev46-browsecomp-correction)

Claude Sonnet 4.6 的 BrowseComp · 1,266 题;单代理与多代理按协议区分 官方出处。见《Claude Sonnet 4.6 System Card》Changelog, PDF p.2 (printed p.1)。证据 ev46-browsecomp-correction。

原文与定位

Claude Sonnet 4.6 · BrowseComp(1,266 题;单代理与多代理按协议区分),No thinking, max effort配置,Accuracy 74.72%(具体成绩公布时间未确认)。见《Claude Sonnet 4.6 System Card》Changelog, PDF p.2 (printed p.1)。 查看成绩条件

Claude Sonnet 4.6 · BrowseComp(1,266 题;单代理与多代理按协议区分),No thinking, max effort配置,Accuracy 74.01%(成绩公布 2026-03-06)。见《Claude Sonnet 4.6 System Card》Changelog, PDF p.2 (printed p.1)。 查看成绩条件

Section 2.20.1.1 and fig. 2.20.1.1.A: Updated highest single-agent BrowseComp score (74.72% -> 74.01%). Improved pipeline flagged 9 additional instances of unintended solutions to the problems that our original pipeline had missed. Instead of re-running the evaluation, we marked those 9 solutions as incorrect. Only the best-scoring single-agent config was re-verified.
官方材料
Claude Sonnet 4.6 System Card · PDF 第 2 页
定位
附件 · Changelog, PDF p.2 (printed p.1)
材料版本
PDF with changelog dated 2026-03-06; 135 pages; collected 2026-11-10
版本公布
未确认
披露时间
· 后补披露
采集时间
历史依据
暂无可靠历史版本依据
引用文本

Claude Sonnet 4.6 · BrowseComp(1,266 题;单代理与多代理按协议区分),No thinking, max effort配置,Accuracy 74.72%(具体成绩公布时间未确认);Claude Sonnet 4.6 · BrowseComp(1,266 题;单代理与多代理按协议区分),No thinking, max effort配置,Accuracy 74.01%(成绩公布 2026-03-06)。官方出处 Claude Sonnet 4.6 System Card(附件,PDF with changelog dated 2026-03-06; 135 pages; collected 2026-11-10)Changelog, PDF p.2 (printed p.1),https://www.anthropic.com/claude-sonnet-4-6-system-card#page=2。Benchlog https://benchlog.lekshome.cc/evidence/claude-sonnet-4-6/ev46-browsecomp-correction/

关联成绩与条件

Claude Sonnet 4.6 · BrowseComp 成绩与配置
指标与成绩配置与评测条件证据关系
Accuracy74.72%原文 74.72% · 单位 %
公布 未确认

已更正 · 74.01%。System Card changelog dated 2026-03-06 states that an improved cheating-detection pipeline flagged nine additional unintended solutions; the best-scoring single-agent configuration was re-verified, not rerun.

No thinking, max effort
设置
Thinking disabled; max effort.
未知项
Sampling settings are not specified in every evaluation.
工具
工具条件未记录
推理预算
max
框架
未记录
协议
BrowseComp single-agent best configuration, 1,266 questions.
条件
Historical result recorded in the current System Card changelog before the 2026-03-06 correction. Main article's summary image still shows the rounded 74.7%. No rerun was performed for the single-agent correction.

直接支持该成绩

Accuracy74.01%原文 74.01% · 单位 %
公布

更正记录 · 替代 74.72%。System Card changelog dated 2026-03-06 states that an improved cheating-detection pipeline flagged nine additional unintended solutions; the best-scoring single-agent configuration was re-verified, not rerun.

No thinking, max effort
设置
Thinking disabled; max effort.
未知项
Sampling settings are not specified in every evaluation.
工具
允许工具
推理预算
max
框架
未记录
协议
BrowseComp single-agent best configuration, 1,266 questions.
条件
Web search, web fetch, programmatic tool calling, context compaction triggered at 50K tokens, 10M total-token limit, max reasoning effort and no thinking.

直接支持该成绩

在排名中查看此成绩