Claude Sonnet 4.6 · MedCalc-Bench(Verified dataset) 86.24% · 附件证据(ev46-medcalc)

Claude Sonnet 4.6 的 MedCalc-Bench · Verified dataset 官方出处。Adaptive thinking, max effort · 86.24%。见《Claude Sonnet 4.6 System Card》§2.21.2, PDF p.52。证据 ev46-medcalc。

原文与定位

Claude Sonnet 4.6 · MedCalc-Bench(Verified dataset),Adaptive thinking, max effort配置,Accuracy 86.24%(具体成绩公布时间未确认)。见《Claude Sonnet 4.6 System Card》§2.21.2, PDF p.52。 查看成绩条件

Claude Sonnet 4.6 achieved 86.24% accuracy, slightly outperforming Claude Opus 4.6 (85.24%), while both 4.6 models demonstrated improvements in patient note interpretation and medical calculation accuracy compared to our previous 4.5 models.
官方材料
Claude Sonnet 4.6 System Card · PDF 第 52 页
定位
附件 · §2.21.2, PDF p.52
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
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 Sonnet 4.6 · MedCalc-Bench(Verified dataset),Adaptive thinking, max effort配置,Accuracy 86.24%(具体成绩公布时间未确认)。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)§2.21.2, PDF p.52,https://www.anthropic.com/claude-sonnet-4-6-system-card#page=52。Benchlog https://benchlog.lekshome.cc/evidence/claude-sonnet-4-6/ev46-medcalc/

关联成绩与条件

Claude Sonnet 4.6 · MedCalc-Bench 成绩与配置
指标与成绩配置与评测条件证据关系
Accuracy86.24%原文 86.24% · 单位 %
公布 未确认
Adaptive thinking, max effort
设置
Adaptive thinking enabled; max effort; default temperature and top_p unless an evaluation specifies otherwise.
未知项
Context window is evaluation-dependent; Table 2.1.A states it does not exceed 1M tokens.
工具
工具条件未记录
推理预算
max
框架
未记录
协议
MedCalc-Bench Verified dataset; five-run average; multi-turn agentic loop with Python REPL.
条件
Adaptive thinking and max effort; instructed to write and execute Python code for calculations.

直接支持该成绩

在排名中查看此成绩