ResearchDatasetsAlphaNLIAlphaNLIFollowBenchmarksTaskDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyTaskDatasetSOTA resultTrendResultsAbductive Natural Language InferenceAlphaNLI n=300 per condition (test)95Accuracy21Multi-agent debateAlphaNLI93.7Accuracy15