ResearchTasksGRPO TrainingFor AI assistantsMarkdownContent indexGRPO TrainingFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedGSM8K synthetic variants short-prompt (val)DualKV26.5Peak Memory (GB)8May 18, 2026