problematic is your second table. it is wider than column width. it is posible to make narrower, however content of table is unknown, so the obtained result probably in real document twill not work.

as you can see, i change table layout. to my taste has now better look-out. for this i add two new packages: booktabs for rules and siunitx for S column type. i also assume that content of column header is the widest cells contents.

\documentclass[10pt,twocolumn,letterpaper]{article} \usepackage{siunitx} \usepackage{array, booktabs} \newcommand\mc[1]{\multicolumn{1}{>{\bfseries}c}{#1}} \usepackage{lipsum} \begin{document} \section{Results} \label{Sec:results} \lipsum[22] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Table \begin{table}[htb] \centering \begin{tabular}{l l S[table-format=2.1]} \toprule \mc{Method} & \mc{Input} & \mc{Accuracy} \\ \midrule 3DShapeNets & volumetric & 84.7 \\ Voxnet & volumetric & 85.9 \\ \midrule PointNet & points & 89.2 \\ PointNet++ & points & 90.7 \\ Kd-network & points & {91.8 / 90.6} \\ \midrule Ours & points & {?} \\ \bottomrule \end{tabular} \caption{Classification Accuracy on ModelNet40 dataset} \label{table:ClassificationAcc} \end{table} \lipsum[44] \begin{table}[htb] \centering \begin{tabular}{ l l l l} \toprule \mc{Method} & \mc{Data} & \mc{Permutations} & \mc{Points} \\ \midrule PointNet & & & \\ PointNet++ & & & \\ Kd-Network & & & \\ \bottomrule \end{tabular} \caption{Challenges summary } \label{table:ChallangeComp} \end{table} \lipsum \subsection{Architecture Design Analysis} \label{SubSec:Arch_Dn_Analysis} \subsection{Visualizations} \label{SubSec:Vis} \end{document} 

above mwe gives no error, no warning, no bad boxes.