ResearchTasksMulti-step Reasoning over Code DependenciesFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedFinQA hardSGKR65.56Accuracy6Apr 14, 2026ConvFinQA hardSGKR74.67Accuracy (Multi-step Reasoning)6Apr 14, 2026DABstep hardSGKR24.34Accuracy6Apr 14, 2026