ResearchDatasetsKoalaKoalaFollowBenchmarksTaskDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyTaskDatasetSOTA resultTrendResultsPreference AlignmentKoala196Wins (Count)14Instruction FollowingKoala low-resource58.3Win Rate (bn)7Overrefusal EvaluationKoala4.44Refusal Rate6Camera-controlled video generationKoala0.637RotErr6Personalized LLM response generationKoala (test)88Win Rate (Reward Model)3Preference AlignmentKoala77.75Win Rate (Reward Model)3