ResearchDatasetsHeld-out capabilityFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsCapability RetentionHeld-out capability AIME-2024, MATH-500, GSM8K, HumanEval, MMLU (test)63.5AIME-2024 Score39General capability retentionHeld-out capability average82.6Unweighted Average Accuracy21
Capability RetentionHeld-out capability AIME-2024, MATH-500, GSM8K, HumanEval, MMLU (test)63.5AIME-2024 Score39