The paper [arXiv:2610.23033] studies a spiked Wigner model in which a rank-one signal is corrupted by noise, with both signal and noise having their own sparsity structures. The authors ask two questions: when can an observer tell that a signal is present (distinguishability), and when can the signal be reconstructed (recovery)?

According to the abstract, the analysis covers a variety of algorithms and parameter regimes, and treats the sparsity of the signal and the sparsity of the noise separately. This doubly sparse setting is more general than models that assume sparsity on only one side.

The available abstract does not include specific results or thresholds, so the detailed findings are not yet summarized here. The paper is posted on arXiv under identifier 2610.23033.