Share your thoughts, 1 month free Claude Pro on usSee more

Interactive Environment Task Completion on ALFWorld (Unseen)

91.8Average Reward

EAGLET + GiGPO

Updated 15d ago

Evaluation Results

Method	Links
EAGLET + GiGPO 2025.10		91.8
EAGLET + GPT-5 2025.10		90.7
BPO 2025.08		89.6
GiGPO 2025.10		88.6
MPO + GiGPO 2025.10		88.1
GPT-5 2025.10		83.6
MPO + GPT-5 2025.10		83.6
EAGLET + GPT-4.1 2025.10		83.6
EAGLET + AgentTuning 2025.10		83.4
EAGLET + ETO 2025.10		83.2
MPO + AgentTuning 2025.10		81.3
MPO + GPT-4.1 2025.10		79.1
MPO + ETO 2025.10		79.1
MPO 2025.08		78.4
WKM 2025.10		78.2
ETO 2025.10		76.4
Qwen-2.5-7B-Instruct 2025.08		76.1
KnowAgent 2025.10		74.9
GPT-4.1 2025.10		72.4
AgentTuning 2025.10		71.6
SFT 2025.08		71.6
ETO 2025.08		71.6
DeepSeek-V3.1-Think 2025.10		69.4
Qwen-3-Thinking 2025.08		67.2
o3-mini 2025.08		62.7
EAGLET + Llama-3.1-8B-Instruct 2025.10		55.2
Deepseek-R1 2025.08		53.7
MPO + Llama-3.1-8B-Instruct 2025.10		52.2
Llama-3.1-8B-Instruct 2025.08		40.3
DeepSeek-V3.1-Non-Think 2025.10		37.3
Llama-3.1-8B-Instruct 2025.10		28.4