datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
android-controlAndroidControlParsedWithImages-20kandroid_controlAndroidControlandroidcontrol-rf-sft-v1AndroidControlParsed-20kandroid-control-1kAndroidControlParsedWithImages-20k-TESTONLYandroid-control
Android Control Test Set
Notes
NOTE i uploaded it so it easier for running evaluation on it
I am not responsible if you use it in training :<
I assume most models train on train-set/val-set but not test-set
I assume that the ids here are episode_id
https://console.cloud.google.com/storage/browser/_details/gresearch/android_control/splits.json , so I download data using smolagents
https://huggingface.co/datasets/smolagents/android-control and iterate and filter ids to… See the full description on the dataset page: https://huggingface.co/datasets/aliaagheis/android-control.AndroidControlParsedWithImages-20k-maxnshot2AndroidControl_debugAndroidControl_300samples_qwen2_5vlAndroidControl_300_samples_v2AndroidControl_700samples_qwen2_5vl_intention_refinedandroidcontrol-curated-highAndroidControl_700samples_qwen2_5vl_with_history_2imagesAndroidControl_1000_samplesAndroidControl_700_samples_v2AndroidControl_700samples_qwen2_5vlAndroidControl_700samples_qwen2_5vl_filteredAndroidControl_700samples_qwen2_5vl_with_history_2actionsAndroidControl_3500samples_qwen2_5vlAndroidControl_1500samples_qwen2_5vlAndroidControl_700samples_qwen2_5vl_filtered_balancedAndroidControl_700samples_qwen2_5vl_3500AndroidControl_700samples_qwen2_5vl_intentionAndroidControl_700samples_qwen2_5vl_filtered_balanced_trajectoryAndroidControl_700samples_qwen2_5vl_with_history_2images_refinedAndroidControl_1000_samples_v2AndroidControl_700samples_qwen2_5vl_with_future_image_operation
