Web Task Completion on MiniWoB++ Held-out Tasks (test)
70.4Success RateClaude-3.5-Sonnet
Evaluation Results
| Method | Links | |
|---|---|---|
| Claude-3.5-SonnetBackbone Model=Claude-3.5-Sonnet, Training Protocol=Instruct2025.07 | 70.4 | |
| GPT-4o-MiniBackbone Model=GPT-4o-Mini, Training Protocol=Instruct2025.07 | 66.1 | |
| o1-MiniBackbone Model=o1-Mini, Training Protocol=Instruct2025.07 | 66.1 | |
| Llama-3.1-70B-InstructBackbone Model=Llama-3.1-70B, Training Protocol=Instruct2025.07 | 65.2 | |
| Llama-3.1-405B-InstructBackbone Model=Llama-3.1-405B, Training Protocol=Instruct2025.07 | 65.2 | |
| GPT-4oBackbone Model=GPT-4o, Training Protocol=Instruct2025.07 | 64.3 | |
| Llama-3.1-8B SFT+RL (Ours)Backbone Model=Llama-3.1-8B, Training Protocol=SFT+RL2025.07 | 63.3 | |
| Llama-3.3-70B Instruct (Teacher)Backbone Model=Llama-3.3-70B, Training Protocol=Instruct, Model Role=Teacher2025.07 | 61.9 | |
| Qwen-2.5-7B SFT+RL (Ours)Backbone Model=Qwen-2.5-7B, Training Protocol=SFT+RL2025.07 | 61.5 | |
| Qwen-2.5-72B Instruct (Teacher)Backbone Model=Qwen-2.5-72B, Training Protocol=Instruct, Model Role=Teacher2025.07 | 59 | |
| Llama-3.1-8B SFT (Ours)Backbone Model=Llama-3.1-8B, Training Protocol=SFT2025.07 | 56.7 | |
| Qwen-2.5-7B SFT (Ours)Backbone Model=Qwen-2.5-7B, Training Protocol=SFT2025.07 | 56.5 | |
| Qwen-2.5-7B RL (Ours)Backbone Model=Qwen-2.5-7B, Training Protocol=RL2025.07 | 53 | |
| Llama-3.1-8B RL (Ours)Backbone Model=Llama-3.1-8B, Training Protocol=RL2025.07 | 43.5 | |
| Qwen-2.5-7B Instruct (Student)Backbone Model=Qwen-2.5-7B, Training Protocol=Instruct, Model Role=Student2025.07 | 37 | |
| Llama-3.1-8B Instruct (Student)Backbone Model=Llama-3.1-8B, Training Protocol=Instruct, Model Role=Student2025.07 | 36.4 |