\begin{table}[t]
\caption{Scope and synthesis protocol for the critical narrative review.}\label{tab:review_protocol}
\small
\setlength{\tabcolsep}{3pt}
\begin{tabular}{@{}L{0.20\textwidth}L{0.27\textwidth}L{0.25\textwidth}L{0.18\textwidth}@{}}
\toprule
Literature stream & Selection emphasis & Role in the synthesis & Boundary \\
\midrule
Foundational and theoretical work & Risk correction, transition matrices, identifiability, causal assumptions, and memorization & Establish the assumptions and information limits reused by later methods & Theory without a clear connection to noisy supervision is outside the main scope \\
Representative deep-LNL methods & Methods that introduce a reusable reliability signal or pipeline operation & Compare assumptions, observables, target refinement, representation preservation, and optimization control & Incremental variants are represented selectively rather than exhaustively \\
Benchmarks and evaluation studies & Datasets or protocols that alter the source, structure, or observability of noise & Connect empirical evidence to the noise model and claimed capability & Generic accuracy testbeds are not included solely for completeness \\
Cross-domain and emerging settings & Medical, graph, federated, multimodal, pre-training, preference, and LLM-generated labels & Pressure-test whether the framework survives changes in modality, label space, and unit of noise & Emerging work is treated as directional evidence when consensus is not established \\
\botrule
\end{tabular}
\end{table}