ResearchTasksSparse Autoencoder Feature DecompositionFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedGemma-2-2B layer 12 activationsMatryoshka SAE with rank-r linear bottleneck82Explained Variance (EV)4Jun 15, 2026
Gemma-2-2B layer 12 activationsMatryoshka SAE with rank-r linear bottleneck82Explained Variance (EV)4Jun 15, 2026