Web Agent Tasks on Mind2Web Cross-Task
53.2Step Success RateMINDACT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MINDACTBackbone=Flan-T5-large2024.02 | 53.2 | 59 | 80.6 | 26 | — | |
| MindActw/o HTML=false2024.01 | 52 | 55.1 | 75.7 | — | — | |
| MindActrequired HTML text inputs=true2024.11 | 52 | 55.1 | 75.7 | — | — | |
| Self-MAPBackbone=Flan-T5-large2024.02 | 51.7 | 58.1 | 80.5 | 26.6 | — | |
| MINDACT + FixedBackbone=Flan-T5-large2024.02 | 51.3 | 58 | 79.7 | 26.4 | — | |
| SynapseBackbone=Flan-T5-large2024.02 | 50 | 57.5 | 82 | 23.2 | — | |
| MINDACT + CARBackbone=Flan-T5-large2024.02 | 47.8 | 54.5 | 79.5 | 19.8 | — | |
| Self-MAPBackbone=Flan-T5-base2024.02 | 47.1 | 56.2 | 82.5 | 24.7 | — | |
| ReasoningBankBackbone=Gemini-2.5-pro2025.09 | 45.6 | 53.6 | 62.7 | 5.1 | — | |
| SKILL.nbZero-shot=true2026.06 | 45 | 52 | 63.2 | 4.7 | — | |
| ReasoningBankBackbone=Gemini-2.5-flash2025.09 | 44.9 | 52.1 | 60.4 | 4.8 | — | |
| ReasoningBankZero-shot=true2026.06 | 44.9 | 52.1 | 60.4 | 4.8 | — | |
| SynapseBackbone=Gemini-2.5-pro2025.09 | 44.7 | 50.1 | 61 | 3.6 | — | |
| No MemoryBackbone=Gemini-2.5-pro2025.09 | 44.4 | 49.3 | 60.2 | 3.5 | — | |
| AWMBackbone=Gemini-2.5-pro2025.09 | 44.4 | 48.6 | 61.2 | 3.7 | — | |
| AWMonlineZero-shot=true2026.06 | 43.6 | 50 | 56.4 | 4 | — | |
| Magma-8BBackbone=LLaMA3 [92], DoM Tree=true2025.02 | 43.4 | 54.8 | 79.7 | — | — | |
| MINDACT + FixedBackbone=Flan-T5-base2024.02 | 42.6 | 51 | 80.8 | 18.4 | — | |
| SynapseBackbone=Flan-T5-base2024.02 | 41.9 | 49.6 | 79.9 | 18.4 | — | |
| MINDACT + CARBackbone=Flan-T5-base2024.02 | 41.4 | 47.8 | 78.8 | 16.1 | — | |
| SynapseBackbone=Gemini-2.5-flash2025.09 | 41.2 | 47 | 59.5 | 3.5 | — | |
| AWMBackbone=Gemini-2.5-flash2025.09 | 41 | 46.3 | 56.1 | 3.5 | — | |
| No MemoryBackbone=Gemini-2.5-flash2025.09 | 40.3 | 46 | 59.1 | 3.3 | — | |
| CodeActZero-shot=true2026.06 | 40.3 | 46 | 59.1 | 3.3 | — | |
| SeeAct (GPT-4V)Backbone=GPT-4V [99], Image=true2025.02 | 40.2 | 46.4 | 73.4 | — | — | |
| OmniParser2024.11 | 39.4 | 42.4 | 87.6 | — | — | |
| GPT-4V-OmniParserBackbone=GPT-4V [99], DoM Tree=true, Image=true2025.02 | 39.4 | 42.4 | 87.6 | — | — | |
| Qwen2-UIXBackbone=Qwen2 [124]2025.02 | 38.2 | 43.4 | — | — | — | |
| ShowUIFT?=true2024.11 | 37.2 | 39.9 | 88.6 | — | — | |
| ShowUI †FT?=true, Action history only=true2024.11 | 36.9 | 39.7 | 88 | — | — | |
| MINDACTBackbone=Flan-T5-base2024.02 | 36.6 | 43.2 | 79.1 | 14.2 | — | |
| GPT-4close-source=true2024.11 | 36.3 | 41.6 | 60.6 | — | — | |
| GPT-4w/o HTML=false2024.01 | 36.2 | 41.6 | 60.6 | — | — | |
| GPT-4-MindActBackbone=GPT-4 [98], Image=true2025.02 | 36.2 | 41.6 | 60.6 | — | — | |
| Qwen2-VL-2BFT?=true2024.11 | 33.2 | 37.7 | 86.4 | — | — | |
| CogAgentBackbone=CogVLM [118]2025.02 | 26.9 | 30.2 | — | — | — | |
| DAVELanguage Model Backbone=Llama-3.2-3B-Instruct2025.12 | 26.1 | 30.8 | — | — | — | |
| SeeClickw/o HTML=true2024.01 | 25.5 | 28.3 | 87 | — | — | |
| SeeClickFT?=true2024.11 | 25.5 | 28.3 | 87 | — | — | |
| SeeClickBackbone=Qwen-VL [3]2025.02 | 25.5 | 28.3 | 87 | — | — | |
| MiniCPM-V-GUIBackbone=MiniCPM-V [128], DoM Tree=true2025.02 | 20.8 | 23.8 | 86.8 | — | — | |
| SeeAct (Gemini-Pro)Backbone=Gemini-Pro [36], Image=true2025.02 | 19.6 | 21.5 | 67.7 | — | — | |
| DolphinLanguage Model Backbone=Llama-3.2-3B-Instruct2025.12 | 19.6 | 24 | — | — | — | |
| ShowUI-ZSFT?=false2024.11 | 18.6 | 21.4 | 85.2 | — | — | |
| CogAgentFT?=true2024.11 | 17.6 | 22.4 | 53 | — | — | |
| MindAct (gen)w/o HTML=false2024.01 | 17.5 | 20.2 | 52 | — | — | |
| GPT-3.5-Turbow/o HTML=false2024.01 | 17.4 | 20.3 | 56.6 | — | — | |
| SigLIP 2Language Model Backbone=Llama-3.2-3B-Instruct2025.12 | 17.3 | 20.6 | — | — | — | |
| MAELanguage Model Backbone=Llama-3.2-3B-Instruct2025.12 | 16.8 | 21 | — | — | — | |
| Fuyu-8B-GUIBackbone=Fuyu-8B [4], DoM Tree=true2025.02 | 15.6 | 19.1 | 86.1 | — | — | |
| Qwen-VLw/o HTML=true2024.01 | 13.3 | 15.9 | 86.7 | — | — | |
| Qwen-VLFT?=true2024.11 | 13.3 | 15.9 | 86.7 | — | — | |
| Qwen-VLBackbone=Qwen-VL [3]2025.02 | 13.3 | 15.9 | 86.7 | — | — | |
| Pix2StructLanguage Model Backbone=Llama-3.2-3B-Instruct2025.12 | 12.3 | 15.8 | — | — | — | |
| AIMv2Language Model Backbone=Llama-3.2-3B-Instruct2025.12 | 11.6 | 14.7 | — | — | — | |
| DinoV2Language Model Backbone=Llama-3.2-3B-Instruct2025.12 | 11.4 | 13.5 | — | — | — | |
| MiniCPM-V+Backbone=MiniCPM-V [128], DoM Tree=true2025.02 | 8.5 | 11 | 85.6 | — | — | |
| Fuyu-8B+Backbone=Fuyu-8B [4], DoM Tree=true2025.02 | 6.6 | 8.3 | 83.9 | — | — | |
| DiTLanguage Model Backbone=Llama-3.2-3B-Instruct2025.12 | 4.1 | 5.8 | — | — | — | |
| Evolving-RLSkill injection=true2026.05 | 2.38 | — | — | — | 41.99 | |
| MINDACTBackbone=GPT-3.52024.02 | 1.9 | 4.3 | 27.6 | 1 | — | |
| Evolving-RLSkill injection=false2026.05 | 1.59 | — | — | — | 35.32 | |
| GRPOSkill injection=false2026.05 | 1.51 | — | — | — | 28.79 | |
| GRPOSkill injection=true2026.05 | 0.87 | — | — | — | 28.98 | |
| Base ModelSkill injection=true2026.05 | 0.4 | — | — | — | 14.92 | |
| Memento2026.05 | 0.4 | — | — | — | 15.27 | |
| Base ModelSkill injection=false2026.05 | 0.24 | — | — | — | 12.57 | |
| ReasoningBank2026.05 | 0 | — | — | — | 11.84 | |
| DeBERTaBackbone=DeBERTa2024.02 | — | 36.8 | — | — | — | |
| GPT-4VBackbone=GPT-4V [99], Image=true2025.02 | — | — | 20.3 | — | — |