List your company
ResearchFormerly WizWand

Medical Reasoning on Medical Reasoning

59.7pass@1FedAvg-GRPO
48.7851.61554.4557.285Apr 14, 2026
Updated 6mo ago

Evaluation results

MethodLinks
FedAvg-GRPOLocal steps (τ)=10, Model=Llama3.2-3B-Instruct, Aggregation method=Balanced method, Swap period=22026.0459.7
FedAvg-PubSwapLocal steps (τ)=40, Model=Llama3.2-3B-Instruct, Aggregation method=Balanced method, Swap period=22026.0459.5
FedAvg-GRPOLocal steps (τ)=40, Model=Llama3.2-3B-Instruct, Aggregation method=Balanced method, Swap period=22026.0458.9
FedAvg-PubSwapLocal steps (τ)=90, Model=Llama3.2-3B-Instruct, Aggregation method=Balanced method, Swap period=22026.0458.5
FedAvg-PubSwapLocal steps (τ)=10, Model=Llama3.2-3B-Instruct, Aggregation method=Balanced method, Swap period=22026.0458.3
FedAvg-PubSwapLocal steps (τ)=120, Model=Llama3.2-3B-Instruct, Aggregation method=Balanced method, Swap period=22026.0458.1
FedAvg-GRPOLocal steps (τ)=90, Model=Llama3.2-3B-Instruct, Aggregation method=Balanced method, Swap period=22026.0457.9
FedAvg-GRPOLocal steps (τ)=120, Model=Llama3.2-3B-Instruct, Aggregation method=Balanced method, Swap period=22026.0457.7
Base modelLocal steps (τ)=10, Model=Llama3.2-3B-Instruct, Aggregation method=Balanced method, Swap period=22026.0449.2
Base modelLocal steps (τ)=40, Model=Llama3.2-3B-Instruct, Aggregation method=Balanced method, Swap period=22026.0449.2
Base modelLocal steps (τ)=90, Model=Llama3.2-3B-Instruct, Aggregation method=Balanced method, Swap period=22026.0449.2
Base modelLocal steps (τ)=120, Model=Llama3.2-3B-Instruct, Aggregation method=Balanced method, Swap period=22026.0449.2