Claude Haiku 5.5 · WANDR · 500-task wide research benchmark · 成绩与配置

Claude Haiku 5.5 的 WANDR · 500-task wide research benchmark 记录。本次发布有 5 条成绩记录,含 5 条精确值(可含历史记录),附 1 条来源定位。

成绩与评测条件

Claude Haiku 5.5 · WANDR 成绩与配置
指标与成绩配置与评测条件出处
Soft F13.5%原文 3.5% · 单位 %
公布 未确认
System Card effort / low
设置
Adaptive thinking; low effort, explicitly labeled in the source; benchmark-specific protocol is preserved in each result's conditions.
未知项
Sampling and trial counts vary by evaluation and are recorded per result.
工具
允许工具
推理预算
low
框架
未记录
协议
WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.
条件
low effort. Anthropic used a frozen offline index and Claude Opus 4.6 judge; tool/grading differences mean results are not directly comparable to Perplexity’s. Labeled figure value.
在排名中查看此成绩
Soft F112.9%原文 12.9% · 单位 %
公布 未确认
System Card effort / medium
设置
Adaptive thinking; medium effort, explicitly labeled in the source; benchmark-specific protocol is preserved in each result's conditions.
未知项
Sampling and trial counts vary by evaluation and are recorded per result.
工具
允许工具
推理预算
medium
框架
未记录
协议
WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.
条件
medium effort. Anthropic used a frozen offline index and Claude Opus 4.6 judge; tool/grading differences mean results are not directly comparable to Perplexity’s. Labeled figure value.
在排名中查看此成绩
Soft F137.3%原文 37.3% · 单位 %
公布 未确认
System Card effort / high
设置
Adaptive thinking; high effort, explicitly labeled in the source; benchmark-specific protocol is preserved in each result's conditions.
未知项
Sampling and trial counts vary by evaluation and are recorded per result.
工具
允许工具
推理预算
high
框架
未记录
协议
WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.
条件
high effort. Anthropic used a frozen offline index and Claude Opus 4.6 judge; tool/grading differences mean results are not directly comparable to Perplexity’s. Labeled figure value.
在排名中查看此成绩
Soft F145.9%原文 45.9% · 单位 %
公布 未确认
System Card effort / xhigh
设置
Adaptive thinking; xhigh effort, explicitly labeled in the source; benchmark-specific protocol is preserved in each result's conditions.
未知项
Sampling and trial counts vary by evaluation and are recorded per result.
工具
允许工具
推理预算
xhigh
框架
未记录
协议
WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.
条件
xhigh effort. Anthropic used a frozen offline index and Claude Opus 4.6 judge; tool/grading differences mean results are not directly comparable to Perplexity’s. Labeled figure value.
在排名中查看此成绩
Soft F149.9%原文 49.9% · 单位 %
公布 未确认
System Card effort / max
设置
Adaptive thinking; max effort, explicitly labeled in the source; benchmark-specific protocol is preserved in each result's conditions.
未知项
Sampling and trial counts vary by evaluation and are recorded per result.
工具
允许工具
推理预算
max
框架
未记录
协议
WANDR; 500 wide-research tasks; web search, web fetch, programmatic tool calling, code execution; 980k-token budget; soft F1 averaged across tasks.
条件
max effort. Anthropic used a frozen offline index and Claude Opus 4.6 judge; tool/grading differences mean results are not directly comparable to Perplexity’s. Labeled figure value.
在排名中查看此成绩

出处与披露

测试版本与指标

别名 Wide ANd Deep Research

  • Soft F1 · % · 越高越好