ResearchTasksLogit Divergence AnalysisLogit Divergence AnalysisFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedPhysical reasoning domain (calibration set)µCRASP0.92Mean KL Divergence (nats)5May 26, 2026