Interactive Shopping on WebShop (test)
0.8Average Rewardturn-PPO
Evaluation Results
| Method | Links | |
|---|---|---|
| turn-PPOBackbone=Qwen3-1.7B, Thinking Mode=no think2025.12 | 0.8 | |
| GRPOBackbone=Qwen3-1.7B, Thinking Mode=no think2025.12 | 0.78 | |
| Token-PPOBackbone=Qwen3-1.7B, Thinking Mode=no think2025.12 | 0.77 | |
| turn-PPOBackbone=Qwen2.5-3B2025.12 | 0.75 | |
| Token-PPOBackbone=Qwen2.5-3B2025.12 | 0.73 | |
| GRPOBackbone=Qwen2.5-3B2025.12 | 0.72 | |
| turn-PPOBackbone=Qwen3-1.7B, Thinking Mode=think2025.12 | 0.55 | |
| Token-PPOBackbone=Qwen3-1.7B, Thinking Mode=think2025.12 | 0.54 |