\begin{table}[t]
\caption{Peer-reviewed 2026 evidence in noisy-label learning.}\label{tab:recent2026}
\small
\begin{tabular}{@{}L{0.22\textwidth}L{0.16\textwidth}L{0.24\textwidth}L{0.28\textwidth}@{}}
\toprule
Paper & Evidence status & Main setting & Role in this review \\
\midrule
Label-noise SGD dynamics \citep{zhang2026labelsgd} & Published, AAAI 2026 & Theory and optimization & Explains how label noise can drive rich-regime dynamics and connects noisy-label training to SAM-like optimization \\
Jump-teaching \citep{ji2026jumpteaching} & Published, AAAI 2026 & Efficient sample selection & Uses temporal disagreement to reduce selection bias without dual networks or extra forward passes \\
DKAF \citep{ning2026dkaf} & Published, AAAI 2026 & VLM noisy-label learning & Shows that VLM priors may create endogenous confirmation bias during noisy-label detection and correction \\
FedRNC \citep{huang2026fedrnc} & Published, AAAI 2026 & Federated class-incremental LNL & Introduces spatio-temporal label misalignment as a federated noisy-supervision failure mode \\
Prototype-guided graph supervision \citep{li2026prototype} & Published, AAAI 2026 & Graph learning & Replaces direct noisy node labels with class-level prototype supervision under sparse labels \\
\botrule
\end{tabular}
\end{table}