1\begin{table}[htbp]\centering\footnotesize2\caption{Preregistered paired contrasts (cluster bootstrap 95\% CI, Holm-adjusted Wilcoxon p)}\label{tab:hypotheses}3\resizebox{\linewidth}{!}{\begin{tabular}{lrrrrrr}\toprule4Condition & n & mean diff & ci low & ci high & holm p & cliffs delta \\\midrule5H1\_D\_vs\_B / clarity & 24.00 & -0.12 & -0.31 & 0.03 & 0.72 & -0.12 \\6H1\_D\_vs\_B / argument\_logic & 24.00 & 0.04 & -0.16 & 0.22 & 1.00 & 0.04 \\7H1\_D\_vs\_B / contribution\_positioning & 24.00 & 0.00 & -0.28 & 0.28 & 1.00 & 0.01 \\8H1\_D\_vs\_B / usefulness & 24.00 & 0.04 & -0.28 & 0.41 & 1.00 & 0.04 \\9H2\_D\_vs\_C / clarity & 24.00 & -0.04 & -0.09 & 0.00 & 0.95 & -0.04 \\10H2\_D\_vs\_C / argument\_logic & 24.00 & -0.12 & -0.22 & 0.03 & 0.72 & -0.12 \\11H2\_D\_vs\_C / contribution\_positioning & 24.00 & 0.12 & -0.03 & 0.22 & 0.63 & 0.13 \\12H2\_D\_vs\_C / usefulness & 24.00 & 0.08 & -0.06 & 0.34 & 0.48 & 0.07 \\13H3\_E\_vs\_D / methodological\_rigor & 24.00 & 0.96 & 0.44 & 1.47 & 0.01 & 0.55 \\14H3\_E\_vs\_D / fidelity & 24.00 & 0.46 & 0.03 & 0.69 & 0.02 & 0.32 \\15H3\_E\_vs\_D / usefulness & 24.00 & -0.33 & -0.78 & 0.06 & 0.06 & -0.24 \\16E\_vs\_A / clarity & 24.00 & 0.04 & -0.25 & 0.69 & 0.74 & -0.11 \\17E\_vs\_A / argument\_logic & 24.00 & 0.46 & 0.09 & 1.31 & 0.46 & 0.24 \\18E\_vs\_A / contribution\_positioning & 24.00 & -0.08 & -0.38 & 0.72 & 0.97 & -0.14 \\19E\_vs\_A / methodological\_rigor & 24.00 & 1.33 & 0.84 & 2.28 & 0.00 & 0.65 \\20E\_vs\_A / fidelity & 24.00 & 0.38 & -0.03 & 1.12 & 0.29 & 0.22 \\21E\_vs\_A / usefulness & 24.00 & -0.17 & -0.56 & 0.44 & 0.94 & -0.21 \\22E\_vs\_B / clarity & 24.00 & -0.21 & -0.47 & -0.06 & 0.13 & -0.19 \\23E\_vs\_B / argument\_logic & 24.00 & 0.33 & 0.09 & 0.69 & 0.25 & 0.27 \\24E\_vs\_B / contribution\_positioning & 24.00 & -0.42 & -0.75 & 0.09 & 0.29 & -0.29 \\25E\_vs\_B / methodological\_rigor & 24.00 & 1.08 & 0.56 & 1.81 & 0.01 & 0.55 \\26E\_vs\_B / fidelity & 24.00 & 0.38 & 0.16 & 0.59 & 0.06 & 0.30 \\27E\_vs\_B / usefulness & 24.00 & -0.29 & -0.66 & 0.03 & 0.23 & -0.21 \\28\bottomrule\end{tabular}}29\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}30\end{table}31