List your company

Human behavior simulation on SHOP-CART (test)

0.2772Exact Action AccuracyShop-R1
-0.007760.066220.14020.21418Jul 23, 2025

Evaluation results

MethodLinks
Shop-R1Model=Qwen-2.5-3B-Instruct2025.070.27720.3640.3128
Shop-R1Model=Qwen-2.5-0.5B-Instruct2025.070.27720.31830.212
Shop-R1Model=Qwen-2.5-1.5B-Instruct2025.070.24110.34540.2919
SFTModel=Qwen-2.5-3B-Instruct2025.070.16760.22250.2452
SFT + RL (Binary)Model=Qwen-2.5-3B-Instruct2025.070.16550.23740.2807
SFTModel=Qwen-2.5-1.5B-Instruct2025.070.10860.23580.2902
SFTModel=Qwen-2.5-0.5B-Instruct2025.070.0990.17720.2161
Zero-shot promptingModel=Qwen-2.5-0.5B-Instruct2025.070.06760.12880.1555
RL (Binary)Model=Qwen-2.5-3B-Instruct2025.070.01010.06170.0992
Zero-shot promptingModel=Qwen-2.5-1.5B-Instruct2025.070.00530.03940.0616
Zero-shot promptingModel=Qwen-2.5-3B-Instruct2025.070.00320.15330.1615