always-further/deepfabric
DeepFabric generates synthetic training and evaluation datasets for agentic models, combining topic graph expansion, reasoning traces, constrained decoding and real tool execution in isolated WebAssembly sandboxes.
Awesome AI Agents › Training Datasets
OpenGVLab/GUI-Odyssey
GUI Odyssey is a comprehensive dataset designed for training and evaluating cross-application navigation agents on mobile devices. It includes 7,735 episodes collected from six different mobile devices, covering six types of cross-app navigation tasks, involving 201 distinct applications and 1,400 application combinations. The dataset aims to facilitate research and development in the area of GUI navigation, particularly focusing on the challenges of navigating across multiple apps on mobile platforms. The dataset was created through a meticulous data collection pipeline where human demonstrations were recorded on an Android emulator. These demonstrations capture detailed metadata for each navigation episode, ensuring high-quality and diverse data. The dataset is organized into various splits such as random, task-based, device-based, and app-based splits, allowing for comprehensive evaluation scenarios. Each episode includes unique instructions generated from predefined templates, which guide the navigation tasks. GUI Odyssey is hosted on Huggingface, making it easily accessible for researchers and developers. The repository provides scripts for preprocessing the data and organizing screenshots, facilitating straightforward integration into machine learning workflows. Additionally, the project includes models trained on the dataset, such as the OdysseyAgent, which can be used for benchmarking and further research. The dataset and accompanying code support a wide range of research in mobile GUI navigation, including the development of intelligent agents capable of understanding and performing complex navigation tasks across multiple applications. The project is well-documented, with detailed statistics, data structure explanations, and quick start guides to help users get up and running quickly. Overall, GUI Odyssey represents a significant resource for advancing the state of the art in cross-app navigation on mobile devices, providing a rich dataset and tools to foster innovation in this emerging field.
https://github.com/OpenGVLab/GUI-Odyssey
Also in Training Datasets
Dongping-Chen/GUI-World
GUI-World is a dataset and benchmark for evaluating and advancing multimodal large language models in understanding and interacting with dynamic graphical user interfaces, complemented by GUI-Vid, a fine-tuned VideoLLM for GUI video analysis.
neulab/MultiUI
MultiUI is a large-scale dataset and codebase for training and evaluating models on text-rich visual understanding of webpage user interfaces, featuring 7.3 million samples, pre-trained models, and comprehensive benchmark evaluations.