% tab1_arms.tex -- GENERATED by tables/make_tab1.py; edit the script, not this file.
% pass@1 = solved/731 (Pro) and solved/500 (Verified) from data/report_excerpt.json; deltas from its
% `tests`; teleport-excluded deltas from analysis/stats/stats_fixes.md section e;
% cumulative input: report pages + 算_行为分解.py (see the docstring); Bash-JS: fp_summary.json.
\begin{table*}[t]
  \centering
  \footnotesize
  \setlength{\tabcolsep}{3.5pt}
  \begin{tabular}{@{}lrrrrrrr@{}}
    \toprule
    Arm (base model + corpus dose) & \multicolumn{2}{c}{\passat (\%)} & \multicolumn{2}{c}{$\Delta$ vs.\ native (pp)} & $\Delta$Pro$-$tp & Input & Bash-\JS \\
    \cmidrule(lr){2-3}\cmidrule(lr){4-5}
     & Pro & Ver. & Pro & Ver. & (pp) & (M tok) & to teacher \\
    \midrule
    \nTwoSeven & 50.1 & 68.4 & -- & -- & -- & 3.27 & 0.202 \\
    \sOneTwoEight & 49.7 & 67.8 & $-0.41$ & $-0.60$ & $-1.22$ & 1.69 & 0.069 \\
    \sTwoFiveSix & 52.8 & 68.6 & $+2.74$ & $+0.20$ & $+2.60$ & 1.52 & 0.053 \\
    \sThreeEightOne & 51.0 & 71.2 & $+0.96$ & $+2.80$ & $-0.15$ & 1.75 & 0.050 \\
    \vThirteen & 52.3 & 66.2 & $+2.19$ & $-2.20$ & $+1.98$ & 1.80 & 0.055 \\
    \midrule
    \aBase & 35.4 & 52.0 & -- & -- & -- & 3.40 & 0.278 \\
    \aTwoEp & 44.3 & 55.4 & $+8.89$ & $+3.40$ & $+9.62$ & 2.44 & 0.057 \\
    \midrule
    \aQ & 46.1 & 63.0 & -- & -- & -- & 3.27 & 0.195 \\
    \aV & 48.8 & 63.0 & $+2.74$ & $+0.00$ & $+3.05$ & 2.05 & 0.061 \\
    \midrule
    % repeat-sample twins: kept as rows (PAPER_PLAN v1: 12-arm table); the body uses them as the run-to-run reference.
    \aQt & 47.1 & -- & -- & -- & -- & 3.13 & 0.200 \\
    \aVt & 49.5 & -- & $+2.46$ & -- & $+1.83$ & 2.01 & 0.069 \\
    \midrule
    \qThreeEight & 55.7 & 74.2 & $+5.61$ & $+5.80$ & $+5.04$ & 4.41 & 0.061 \\
    \bottomrule
  \end{tabular}
  \caption{The twelve evaluation arms, named by base model and corpus dose (\dose{v12}{2} = two epochs of \vTwelve; \dose{v13}{1} = one epoch of \vThirteenCorpus, 362 updates). \passat = solved tasks over the pool (731 \SWEPro, 500 \SWEVerified; anomalous trials count as unsolved), one rollout. $\Delta$: percentage points against the line's native arm (\qThreeEight: against \nTwoSeven, the generation jump). $\Delta$Pro$-$tp: Pro without the 76 teleport tasks ($N=655$). Input: median cumulative input tokens per task on Pro, millions. Bash-\JS: Jensen--Shannon distance ($\log_2$) of the Bash head-word distribution to the teacher on Pro. The two $T{=}0.6$ rows are the same weights as the rows above them, run a second time (the run-to-run reference, \S\ref{sec:measurement}). ``--'': not defined for a native arm.}
  \label{tab:arms}
\end{table*}