multimodal benchmark
AITZ_EM
Android-In-The-Zoo (AitZ) benchmark for evaluating autonomous GUI agents on smartphones. Contains 18,643 screen-action pairs with chain-of-action-thought annotations spanning over 70 Android apps. Designed to connect perception (screen layouts and UI elements) with cognition (action decision-making) for natural language-triggered smartphone task completion.
Updated Aug 7, 2026
Published models3
Registry coverage3
MetricScore
EvidenceB
AITZ_EM leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AC | 83.2% | 100.0% | 3 | C | |
| 2 | AC | 83.1% | 50.0% | 3 | C | |
| 3 | AC | 81.9% | 0.0% | 3 | C |
Score distribution
Top published rows on the benchmark's original scale.
AITZ_EM highlights
The top published results on this benchmark's own scale.
What is AITZ_EM?
Definition and scoring fields from the benchmark registry.
Android-In-The-Zoo (AitZ) benchmark for evaluating autonomous GUI agents on smartphones. Contains 18,643 screen-action pairs with chain-of-action-thought annotations spanning over 70 Android apps. Designed to connect perception (screen layouts and UI elements) with cognition (action decision-making) for natural language-triggered smartphone task completion.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- AITZ_EM
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- aitz-em|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about AITZ_EM.
Which model scores highest on AITZ_EM?
Qwen2.5 VL 72B Instruct is currently ranked first with 83.2%.
What does AITZ_EM measure?
Android-In-The-Zoo (AitZ) benchmark for evaluating autonomous GUI agents on smartphones. Contains 18,643 screen-action pairs with chain-of-action-thought annotations spanning over 70 Android apps. Designed to connect perception (screen layouts and UI elements) with cognition (action decision-making) for natural language-triggered smartphone task completion.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
3 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.