Action Prediction on MIND2WEB Cross-Task 1.0
60.6Element AccuracyHTML-T5-XL
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| HTML-T5-XLTrain=SL2023.07 | 60.6 | 81.7 | 57.8 | — | |
| MindActBackbone=Flan-T5-XL2023.06 | 55.1 | 75.7 | 52 | 5.2 | |
| MindAct (Flan-T5-XL)Train=SL, LLM=Flan-T5-XL2023.07 | 55.1 | 75.7 | 52 | — | |
| MindActBackbone=Flan-T5-Large2023.06 | 53.4 | 75.7 | 50.3 | 7.1 | |
| MindActBackbone=Flan-T5-Base2023.06 | 43.6 | 76.8 | 41 | 4 | |
| MindActBackbone=GPT-4, Tasks=50, Candidates=Top-102023.06 | 41.6 | 60.6 | 36.2 | 2 | |
| MindAct (GPT-4)Train=ICL, LLM=GPT-42023.07 | 41.6 | 60.6 | 36.2 | — | |
| Synapse (GPT-3.5)Train=ICL, LLM=GPT-3.52023.07 | 34.4 | — | 30.6 | — | |
| ClassificationBackbone=DeBERTa-Base2023.06 | 26.8 | — | — | — | |
| MindActBackbone=GPT-3.52023.06 | 20.3 | 56.6 | 17.4 | 0.8 | |
| GenerationBackbone=Flan-T5-Base2023.06 | 20.2 | 52 | 17.5 | 0 |