ResearchBenchmarksBehavior Prediction on Carat Top 1000 Users App Usage Dataset (test)Behavior Prediction on Carat Top 1000 Users App Usage Dataset (test)Follow44.7Weighted PrecisionBUA29.30833.30437.341.296Apr 26, 2026Evaluation resultsMethodMethodLinksWeighted PrecisionWeighted RecallOverall ScoreHead Category ScoreMedium Category ScoreTail Category ScoreBUA2026.0444.741.84040.945.126.7CoLLM2026.044036.536.737.741.823.3PITuning2026.0435.235.635.736.242.515.2BehaveGPT2026.0429.931.82130.321.95.2