List your company

Embodied household manipulation on ALFWorld Unseen domain

94Success RateSkillMAS
-0.660823.914648.4973.0654Nov 18, 2025Dec 16, 2025Jan 14, 2026Feb 12, 2026Mar 13, 2026Apr 11, 2026May 10, 2026

Evaluation results

MethodLinks
SkillMASEvaluation Source=Internal evaluation2026.0594
Traj-BootstrapEvaluation Source=Published reference (contextual)2026.0593
CDMemEvaluation Source=Published reference (contextual)2026.0585.8
GRPOFramework=Agent-R1, Backbone=Qwen3-4B, Type=RL-optimized2025.1174.58
RLOOFramework=Agent-R1, Backbone=Qwen3-4B, Type=RL-optimized2025.1173.46
PPOFramework=Agent-R1, Backbone=Qwen3-4B, Type=RL-optimized2025.1172.38
Reinforce++Framework=Agent-R1, Backbone=Qwen3-4B, Type=RL-optimized2025.1169.57
ReActEvaluation Source=Published reference (contextual)2026.0553
Direct LLMEvaluation Source=Published reference (contextual)2026.0543.3
ReActFramework=Agent-R1, Backbone=Qwen3-4B, Type=Training-free baseline2025.112.98