List your company

Long-horizon Decision Making on ALFWorld tool-use (test)

77.8Success RateReAct + FLARE
60.6465.09569.5574.005Jan 29, 2026

Evaluation results

MethodLinks
ReAct + FLARELookahead=Tree, Value Prop.=Backward, Commit=Receding-horizon2026.0177.8
RL Policy (Think)Lookahead=Partial, Value Prop.=Partial, Commit=Full2026.0177.4
ReAct + MCTSLookahead=Tree, Value Prop.=Backward, Commit=Full2026.0173.5
RL Policy (Greedy)Lookahead=None, Value Prop.=None, Commit=Receding-horizon2026.0169.4
ReAct + BeamLookahead=None, Value Prop.=None, Commit=Full2026.0166.7
ReflexionLookahead=None, Value Prop.=None, Commit=Receding-horizon2026.0164.2
ReActLookahead=None, Value Prop.=None, Commit=Receding-horizon2026.0161.3