\begin{table}[t]
\caption{Named comparison with representative prior surveys and adjacent reviews. The comparison concerns organizing emphasis rather than overall quality or comprehensiveness.}\label{tab:survey_positioning}
\footnotesize
\setlength{\tabcolsep}{3pt}
\begin{tabular}{@{}L{0.20\textwidth}L{0.22\textwidth}L{0.25\textwidth}L{0.23\textwidth}@{}}
\toprule
Prior review & Primary scope & Organizing emphasis & Boundary relative to this review \\
\midrule
Fr\'{e}nay and Verleysen \citep{frenay2014classification} & Classical classification with mislabeled training instances & Noise sources, consequences, noise-tolerant learning, label cleansing, and experimental design & Predates the current deep-learning emphasis on memorization, representation dynamics, and foundation-model supervision \\
Cordeiro and Carneiro \citep{cordeiro2020survey} & Deep learning with noisy labels & Taxonomy of robust training strategies and benchmark datasets & Provides a method-family overview; the present review instead connects assumptions, training signals, pipeline dependencies, and evaluation realism \\
Han et al. \citep{han2020representation} & Label-noise representation learning & Historical and theoretical perspectives on learning representations under noise & Closest to our representation emphasis, but not organized around realistic noise provenance and cross-scenario pressure tests \\
Song et al. \citep{song2023survey} & Deep neural networks trained with noisy labels & Five robust-training groups, comparison properties, noise-rate estimation, datasets, and metrics & Offers broad method coverage; our focus is the dependency between formulation, dynamics, pipeline role, and evidence supplied by a benchmark \\
Shin et al. \citep{shin2024cleaning} & Label cleaning & Identification and correction of erroneous labels & Concentrates on data cleaning rather than the full robust-training and representation-learning pipeline \\
Shi et al. \citep{shi2024medical} & Medical image analysis & Domain-specific noise sources, methods, and clinical evaluation issues & Provides medical depth; our review uses medical evidence as one pressure test within a cross-domain framework \\
Song et al. \citep{song2025survey} & Deep LNL and application settings & Method taxonomy and application-oriented overview & Extends application coverage; our synthesis centers on reliability signals, stage interactions, and assumption failures \\
Nazaretyan et al. \citep{nazaretyan2025benchmarking} & Identification of mislabeled tabular data & Empirical comparison of label-error filters and data-dependent recommendations & Benchmarks data-quality detection rather than end-to-end deep LNL, representation dynamics, or target refinement \\
\botrule
\end{tabular}
\end{table}