mirror of
https://github.com/hanzoai/papers.git
synced 2026-10-29 07:34:31 +00:00
29 lines
939 B
TeX
29 lines
939 B
TeX
\section{Benchmark Task Breakdown}
|
|
|
|
\label{app:benchmark}
|
|
|
|
\begin{table}[H]
|
|
\centering
|
|
\begin{tabular}{lcccc}
|
|
\toprule
|
|
Task Category & $N$ & Gen.\ Tool & No Gen.\ Tool & Delta \\
|
|
\midrule
|
|
File manipulation & 85 & 96.5\% & 88.2\% & +8.3\% \\
|
|
API integration & 110 & 94.5\% & 83.6\% & +10.9\% \\
|
|
Code refactoring & 95 & 91.6\% & 81.1\% & +10.5\% \\
|
|
Data transformation & 80 & 93.8\% & 85.0\% & +8.8\% \\
|
|
System configuration & 65 & 92.3\% & 82.0\% & +10.3\% \\
|
|
Documentation & 65 & 89.2\% & 82.0\% & +7.2\% \\
|
|
\midrule
|
|
\textbf{Overall} & \textbf{500} & \textbf{93.1\%} & \textbf{83.8\%} & \textbf{+9.3\%} \\
|
|
\bottomrule
|
|
\end{tabular}
|
|
\caption{Per-category breakdown of the specific-tools advantage.
|
|
The gap is largest for API integration (+10.9\%) and smallest for
|
|
documentation generation (+7.2\%), consistent with the intuition
|
|
that structured, programmatic tasks benefit most from purpose-built
|
|
tools.}
|
|
\label{tab:breakdown}
|
|
\end{table}
|
|
|