\begin{table}[htbp]\centering\footnotesize
\caption{Preregistered paired contrasts (cluster bootstrap 95\% CI, Holm-adjusted Wilcoxon p)}\label{tab:hypotheses}
\begin{tabular}{lrrrrrr}\toprule
Condition & n & mean\_diff & ci\_low & ci\_high & holm\_p & cliffs\_delta \\\midrule
H1\_D\_vs\_B / clarity & 24.00 & 0.04 & -0.41 & 0.28 & 1.00 & 0.01 \\
H1\_D\_vs\_B / argument\_logic & 24.00 & 0.33 & -0.22 & 0.66 & 0.45 & 0.20 \\
H1\_D\_vs\_B / contribution\_positioning & 24.00 & 0.08 & -0.41 & 0.44 & 0.97 & 0.02 \\
H1\_D\_vs\_B / usefulness & 24.00 & 0.17 & -0.50 & 0.69 & 1.00 & 0.01 \\
H2\_D\_vs\_C / clarity & 24.00 & 0.00 & -0.28 & 0.25 & 1.00 & 0.01 \\
H2\_D\_vs\_C / argument\_logic & 24.00 & 0.08 & -0.44 & 0.47 & 1.00 & 0.00 \\
H2\_D\_vs\_C / contribution\_positioning & 24.00 & -0.21 & -0.72 & 0.16 & 0.61 & -0.13 \\
H2\_D\_vs\_C / usefulness & 24.00 & 0.04 & -0.50 & 0.44 & 1.00 & -0.07 \\
H3\_E\_vs\_D / methodological\_rigor & 24.00 & 1.33 & 1.00 & 1.84 & 0.00 & 0.68 \\
H3\_E\_vs\_D / fidelity & 24.00 & 0.62 & 0.31 & 1.03 & 0.00 & 0.36 \\
H3\_E\_vs\_D / usefulness & 24.00 & 0.25 & -0.03 & 0.66 & 0.11 & 0.17 \\
E\_vs\_A / clarity & 24.00 & -0.04 & -0.41 & 0.28 & 0.78 & -0.06 \\
E\_vs\_A / argument\_logic & 24.00 & 0.92 & 0.41 & 1.44 & 0.00 & 0.52 \\
E\_vs\_A / contribution\_positioning & 24.00 & 0.50 & 0.12 & 0.91 & 0.06 & 0.20 \\
E\_vs\_A / methodological\_rigor & 24.00 & 2.04 & 1.44 & 2.62 & 0.00 & 0.85 \\
E\_vs\_A / fidelity & 24.00 & 0.42 & 0.16 & 0.78 & 0.05 & 0.28 \\
E\_vs\_A / usefulness & 24.00 & 0.46 & -0.09 & 1.03 & 0.17 & 0.16 \\
E\_vs\_B / clarity & 24.00 & -0.04 & -0.41 & 0.25 & 0.81 & -0.05 \\
E\_vs\_B / argument\_logic & 24.00 & 0.79 & 0.41 & 1.06 & 0.00 & 0.48 \\
E\_vs\_B / contribution\_positioning & 24.00 & 0.25 & -0.12 & 0.69 & 0.54 & 0.09 \\
E\_vs\_B / methodological\_rigor & 24.00 & 1.67 & 1.12 & 2.19 & 0.00 & 0.74 \\
E\_vs\_B / fidelity & 24.00 & 1.04 & 0.78 & 1.44 & 0.00 & 0.57 \\
E\_vs\_B / usefulness & 24.00 & 0.42 & -0.12 & 0.97 & 0.36 & 0.15 \\
\bottomrule\end{tabular}
\begin{minipage}{0.95\linewidth}\vspace{2pt}\scriptsize Notes: generated by scripts/05\_analyze\_benchmark.py from recorded runs (protocol heldout\_v1); judge = claude-sonnet-5-5 blind to condition; scores 1--7.\end{minipage}
\end{table}