Web Task on Mind2Web (test)
31.3Step-wise Success RateLUMOS-I_Web-13B
Evaluation Results
| Method | Links | |
|---|---|---|
| LUMOS-I_Web-13BType=Open-source, Fine-tuned=true, Backbone Size=13B2023.11 | 31.3 | |
| LUMOS-I_WebType=Open-source, Fine-tuned=true2023.11 | 27.6 | |
| GPT-4Type=GPT/API-based2023.11 | 22.6 | |
| GPT-3.5-turboType=GPT/API-based2023.11 | 15.7 | |
| AgentLM-70BType=Open-source, Fine-tuned=true2023.11 | 13.5 | |
| Koala-13BType=Open-source2023.11 | 6 | |
| WizardLM-30BType=Open-source2023.11 | 3.1 | |
| Baichuan-13B-chatType=Open-source2023.11 | 2.3 |