ResearchTasksResponse Consistency EvaluationFor AI assistantsMarkdownContent indexResponse Consistency EvaluationFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedTPI-Bench Shared utterances between TPI and Janus (test)Kimi-Audio-Instruct-7B94BLEU8Apr 21, 2026