List your company
ResearchFormerly WizWand

Multi-Task Agent Generalization on Five Agent Environments Average (test)

73.2Average Success RateFed-SE
3.20821.37939.5557.721Dec 9, 2025
Updated 8mo ago

Evaluation results

MethodLinks
Fed-SEBase Model=Qwen2.5-7B2025.1273.2
Fed-SEBase Model=Qwen3-8B2025.1270.2
CentralizedBase Model=Qwen2.5-7B2025.1266.6
Fed-SEBase Model=Llama2-7B2025.1266.1
LocalBase Model=Qwen2.5-7B2025.1265.7
LocalBase Model=Qwen3-8B2025.1265.2
Fed-SEBase Model=Qwen2.5-3B2025.1263
FedITBase Model=Qwen2.5-7B2025.1262.7
FedITBase Model=Qwen3-8B2025.1261.6
CentralizedBase Model=Qwen2.5-3B2025.1257.3
FedITBase Model=Qwen2.5-3B2025.1257.3
LocalBase Model=Qwen2.5-3B2025.1256.8
FedITBase Model=Llama2-7B2025.1255.7
Fed-SEBase Model=Qwen3-1.7B2025.1254.3
LocalBase Model=Qwen3-1.7B2025.1250.2
CentralizedBase Model=Llama2-7B2025.1249.3
CentralizedBase Model=Qwen3-8B2025.1248
LocalBase Model=Llama2-7B2025.1247.3
CentralizedBase Model=Qwen3-1.7B2025.1246.8
FedITBase Model=Qwen3-1.7B2025.1246.6
Pre-TrainedBase Model=Qwen2.5-7B2025.1218.6
Pre-TrainedBase Model=Qwen3-1.7B2025.1217.7
Pre-TrainedBase Model=Qwen2.5-3B2025.1213.2
Pre-TrainedBase Model=Qwen3-8B2025.1211.4
Pre-TrainedBase Model=Llama2-7B2025.125.9