diff --git a/paper/submission/CRPTO_ijds_submission.tex b/paper/submission/CRPTO_ijds_submission.tex
index 14d5d62..534ad33 100644
--- a/paper/submission/CRPTO_ijds_submission.tex
+++ b/paper/submission/CRPTO_ijds_submission.tex
@@ -766,6 +766,7 @@ \subsection{Multi-Dataset External Economic Replication}
   \label{fig:external-replication}
 \end{figure}
 
+\FloatBarrier
 The external layer also surfaces a result a single-dataset champion cannot show.
 The signed price of robustness---using the same convention as the Lending Club
 field, $(\mathrm{nonrobust}-\mathrm{robust})/\mathrm{nonrobust}$---is a
@@ -810,6 +811,7 @@ \subsection{Multi-Dataset External Economic Replication}
 \end{figure}
 
 % =====================================================================
+\FloatBarrier
 \section{Robustness and Comparators}\label{sec:robustness}
 
 The first robustness concern is temporal leakage. CRPTO addresses it through
@@ -907,6 +909,7 @@ \subsection{Regret-Auditability Frontier}\label{sec:regret}
 $\Gamma_{\mathrm{CP}}$, an exact funded-set certificate, grade-level audits, and
 model-risk lineage that a credit committee can inspect.
 
+\FloatBarrier
 \subsection{Tail Risk and Distribution Robustness}\label{sec:tail-dist}
 
 Two reviewer questions deserve a body-level answer: what does the champion give up