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

Interactive Agent Task on ScienceWorld Seen (Average Reward)

77.1Average Reward

BPO

Updated 15d ago

Evaluation Results

Method	Links
BPO 2025.08		77.1
MPO 2025.08		71.61
ETO 2025.08		65.69
Deepseek-R1 2025.08		63.96
SFT 2025.08		58.82
o3-mini 2025.08		56.95
Qwen-3-Thinking 2025.08		52.05
Qwen-2.5-7B-Instruct 2025.08		26.68
Llama-3.1-8B-Instruct 2025.08		26.64