ResearchTasksPTPTFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedDifferent-environment out-of-domainGemini 3 Flash83.2Accuracy (OOD PT)14Jun 4, 2026AI2-THOR in-domainBagel + Mixed Training66.7Accuracy14Jun 4, 2026