List your company

Sequential Household Interaction Success on ALFWorld OOD (test)

96Pick Success RateReflexion
0.3225.165074.84May 6, 2026

Evaluation results

MethodLinks
ReflexionSupervision=Access to text env, Agent type=LLM2026.05969410081838891
DEPSSupervision=Access to text env, Agent type=LLM2026.05931005080100076
VIPER BCSupervision=No access to text env, Agent type=VLM+LLM2026.0580776787715372
VIPER BC+RLSupervision=No access to text env, Agent type=VLM+LLM2026.0580777792715375
PRISM BCSupervision=No access to text env, Agent type=VLM+LLM2026.0580778187716277
PRISM BC+RLSupervision=No access to text env, Agent type=VLM+LLM2026.0580838492716780
EMAC+Supervision=Access to text env, Agent type=VLM2026.0579889390907486
Idefics2 8B*Supervision=No access to text env, Agent type=VLM2026.0575777469675069
ReActSupervision=Access to text env, Agent type=LLM2026.0571286562443554
EMMASupervision=Access to text env, Agent type=VLM2026.0571889485836782
RL4VLMSupervision=No access to text env, Agent type=VLM2026.0547141014181821
MiniGPT-4Supervision=No access to text env, Agent type=VLM2026.0541701917616