List your company
ResearchFormerly WizWand

Embodied Agent Task on VirtualHome Unseen

36.9Success RateGRPO w/ EVU
6.01214.03122.0530.069Apr 19, 2026
Updated 6mo ago

Evaluation results

MethodLinks
GRPO w/ EVUMethod Category=Training, Backbone Model=Qwen3-1.7B-Instruct, Base Method=GRPO, EVU Mechanism=Included2026.0436.9
GRPO w/ EVUMethod Category=Training, Backbone Model=Qwen2.5-3B-Instruct, Base Method=GRPO, EVU Mechanism=Included2026.0436
PPO w/ EVUMethod Category=Training, Backbone Model=Qwen2.5-3B-Instruct, Base Method=PPO, EVU Mechanism=Included2026.0435.2
SFT w/ EVUMethod Category=Training, Backbone Model=Qwen2.5-3B-Instruct, Base Method=SFT, EVU Mechanism=Included2026.0434.4
PPO w/ EVUMethod Category=Training, Backbone Model=Qwen3-1.7B-Instruct, Base Method=PPO, EVU Mechanism=Included2026.0428.8
SFT w/ EVUMethod Category=Training, Backbone Model=Qwen3-1.7B-Instruct, Base Method=SFT, EVU Mechanism=Included2026.0425.6
GRPOMethod Category=Training, Backbone Model=Qwen2.5-3B-Instruct, Base Method=GRPO, EVU Mechanism=None2026.0424.8
PPOMethod Category=Training, Backbone Model=Qwen2.5-3B-Instruct, Base Method=PPO, EVU Mechanism=None2026.0423.2
PPOMethod Category=Training, Backbone Model=Qwen3-1.7B-Instruct, Base Method=PPO, EVU Mechanism=None2026.0422.4
SFTMethod Category=Training, Backbone Model=Qwen2.5-3B-Instruct, Base Method=SFT, EVU Mechanism=None2026.0420
GRPOMethod Category=Training, Backbone Model=Qwen3-1.7B-Instruct, Base Method=GRPO, EVU Mechanism=None2026.0419.4
Plan-and-Act w/ EVUMethod Category=Prompting, Backbone Model=DeepSeek V3.2, Base Method=Plan-and-Act, EVU Mechanism=Included2026.0414.4
ReAct w/ EVUMethod Category=Prompting, Backbone Model=DeepSeek V3.2, Base Method=ReAct, EVU Mechanism=Included2026.0413.6
NoThinking w/ EVUMethod Category=Prompting, Backbone Model=DeepSeek V3.2, Base Method=NoThinking, EVU Mechanism=Included2026.0412.8
Plan-and-ActMethod Category=Prompting, Backbone Model=DeepSeek V3.2, Base Method=Plan-and-Act, EVU Mechanism=None2026.0412.8
ReActMethod Category=Prompting, Backbone Model=DeepSeek V3.2, Base Method=ReAct, EVU Mechanism=None2026.0412.8
SFTMethod Category=Training, Backbone Model=Qwen3-1.7B-Instruct, Base Method=SFT, EVU Mechanism=None2026.048
NoThinkingMethod Category=Prompting, Backbone Model=DeepSeek V3.2, Base Method=NoThinking, EVU Mechanism=None2026.047.2