datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
android_controlAndroidControlParsedWithImages-20kandroidcontrol-rf-sft-v1android-control-1kAndroidControlParsedWithImages-20k-TESTONLYAndroidControlParsedWithImages-20k-maxnshot2android_icon_datasetAndroid icons collected from a variety of sources and grouped manually. Beware of the Misc_or_unknown class that contains the weird icons we couldn't classify.
AndroidControl_debugAndroidControl_300_samples_v2AndroidControl_300samples_qwen2_5vlAndroidControl_700samples_qwen2_5vl_intention_refinedAndroidControl_700_samples_v2AndroidControl_1000_samplesAndroidControl_700samples_qwen2_5vlAndroidControl_700samples_qwen2_5vl_filteredthink-invoke-android
Android Phone Agent Training Dataset
Training data for fine-tuning vision-language models to operate an Android phone autonomously.
Overview
This dataset contains successful task trajectories from an AI agent completing real tasks on an Android device. Each example is a single step in a multi-turn interaction where the agent observes the screen, reasons about what to do, and takes an action.
Source model: Gemini 3 pro
Total examples: 1,138
Unique tasks: 64
Format:… See the full description on the dataset page: https://huggingface.co/datasets/fremko/think-invoke-android.AndroidControl_700samples_qwen2_5vl_with_history_2actionsAndroidControl_1000_samples_v2AndroidControl_1500samples_qwen2_5vlAndroidControl_700samples_qwen2_5vl_intentionAndroidControl_3500samples_qwen2_5vlAndroidControl_700samples_qwen2_5vl_with_history_2imagesAndroidControl_700samples_qwen2_5vl_filtered_balancedAndroidControl_700samples_qwen2_5vl_3500AndroidControl_700samples_qwen2_5vl_filtered_balanced_trajectoryAndroidControl_700samples_qwen2_5vl_with_history_2images_refinedAndroidControl_700samples_qwen2_5vl_with_future_image_operationAndroidControl_700samples_qwen2_5vl_with_future_imageAndroidControl_700samples_operation_qwen2_5vl