ResearchBenchmarksFederated Multi-task Fine-tuning on Multi-task Federated Learning Seq, Token, ChoiceFederated Multi-task Fine-tuning on Multi-task Federated Learning Seq, Token, ChoiceFollow483.767Max RewardUCB-DUAL444.82212454.93281465.0435475.15419Aug 13, 2025Evaluation resultsMethodMethodLinksMax RewardAccuracyLatency (s)Energy (J)UCB-DUALscheduler=enabledscheduler=enabled2025.08483.76788.491487.3786,857.511UCB-DUALscheduler=disabledscheduler=disabled2025.08481.85487.802499.9417,021.568HetLoRA2025.08462.44887.294648.3779,615.848HomoLoRA2025.08446.3285.947712.11812,075.024