ResearchTasksNext-token reasoningNext-token reasoningFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedOMNI-MATH Hard (val)LoopRPT38.1Accuracy10Mar 23, 2026OMNI-MATH Medium (val)LoopRPT61.15Accuracy (Next-token Reasoning)10Mar 23, 2026OMNI-MATH Easy (val)LoopRPT76.89Accuracy10Mar 23, 2026