lightning7 commited on
Commit
1b717ff
·
verified ·
1 Parent(s): aae4e3f

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +130 -0
README.md ADDED
@@ -0,0 +1,130 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - zh
4
+ - en
5
+ license: cc-by-nc-4.0
6
+ size_categories:
7
+ - 1K<n<10K
8
+ task_categories:
9
+ - visual-question-answering
10
+ - image-text-to-text
11
+ tags:
12
+ - gui-agent
13
+ - benchmark
14
+ - multimodal
15
+ pretty_name: UniGUI-Bench
16
+ configs:
17
+ - config_name: element_grounding
18
+ data_files:
19
+ - split: test
20
+ path: data/process_eval/element_grounding.json
21
+ field: examples
22
+ - config_name: action_understanding
23
+ data_files:
24
+ - split: test
25
+ path: data/process_eval/action_understanding.json
26
+ field: examples
27
+ - config_name: page_understanding
28
+ data_files:
29
+ - split: test
30
+ path: data/process_eval/page_understanding.json
31
+ field: examples
32
+ - config_name: action_anticipation
33
+ data_files:
34
+ - split: test
35
+ path: data/process_eval/action_anticipation.json
36
+ field: examples
37
+ - config_name: intent_reasoning
38
+ data_files:
39
+ - split: test
40
+ path: data/process_eval/intent_reasoning.json
41
+ field: examples
42
+ - config_name: action_execution
43
+ data_files:
44
+ - split: test
45
+ path: data/process_eval/action_execution.json
46
+ field: examples
47
+ - config_name: dynamic_adaptation
48
+ data_files:
49
+ - split: test
50
+ path: data/process_eval/dynamic_adaptation.json
51
+ field: examples
52
+ - config_name: task_planning
53
+ data_files:
54
+ - split: test
55
+ path: data/process_eval/task_planning.json
56
+ field: examples
57
+ - config_name: task_progress_awareness
58
+ data_files:
59
+ - split: test
60
+ path: data/process_eval/task_progress_awareness.json
61
+ field: examples
62
+ - config_name: task_completion_detection
63
+ data_files:
64
+ - split: test
65
+ path: data/process_eval/task_completion_detection.json
66
+ field: examples
67
+ - config_name: intern_state_memory
68
+ data_files:
69
+ - split: test
70
+ path: data/process_eval/intern_state_memory.json
71
+ field: examples
72
+ ---
73
+
74
+ # UniGUI-Bench
75
+
76
+ A comprehensive benchmark for evaluating GUI agents across 11 capability dimensions.
77
+
78
+ ## Dataset Structure
79
+
80
+ ### System Evaluation
81
+ - **trajectory**: 1,200 complete GUI task trajectories with step-by-step screenshots
82
+ - **gui_system**: 7,108 individual step evaluation samples
83
+
84
+ ### Process Evaluation (11 Abilities, 200 samples each)
85
+
86
+ | Ability | Description |
87
+ |---------|-------------|
88
+ | element_grounding | Locating UI elements given instructions |
89
+ | action_understanding | Understanding what action was performed |
90
+ | page_understanding | Comprehending page content and layout |
91
+ | action_anticipation | Predicting the next action |
92
+ | intent_reasoning | Inferring user intent from actions |
93
+ | action_execution | Executing the correct action |
94
+ | dynamic_adaptation | Adapting to changing interfaces |
95
+ | task_planning | Planning multi-step tasks |
96
+ | task_progress_awareness | Tracking task completion progress |
97
+ | task_completion_detection | Detecting when a task is done |
98
+ | intern_state_memory | Remembering internal states |
99
+
100
+ ## File Structure
101
+
102
+ ```
103
+ ├── data/
104
+ │ ├── system_eval/
105
+ │ │ ├── trajectory.json
106
+ │ │ └── gui_system.json
107
+ │ └── process_eval/
108
+ │ ├── element_grounding.json
109
+ │ ├── action_understanding.json
110
+ │ └── ... (11 ability files)
111
+ ├── screenshots/
112
+ │ ├── tasks/{task_id}/step_XX.png
113
+ │ └── abilities/{ability_name}/
114
+ └── manifest.json
115
+ ```
116
+
117
+ ## Usage
118
+
119
+ ```python
120
+ from datasets import load_dataset
121
+
122
+ # Load a specific ability
123
+ ds = load_dataset("AGI-Eval/UniGUI-Bench", "element_grounding")
124
+
125
+ # Access samples
126
+ for sample in ds["test"]:
127
+ print(sample["input"])
128
+ print(sample["target"])
129
+ print(sample["extra_data"])
130
+ ```