Search Results for "structured text"

...It fills a gap in automation tooling by focusing on mobile-first workflows where traditional browser or desktop-based automation doesn’t work; such as logistics, gig work, field operations, and other industries reliant on phones or tablets. The project works by using Android’s accessibility API to extract structured UI state (as XML) from the device, which is then fed to a large language model (LLM) like OpenAI’s models for decision-making, and actions are executed via the Android Debug Bridge (ADB). This approach bypasses expensive vision-based models and provides faster, cheaper automation with fine-grained interaction capabilities (for example, tapping buttons, typing text, navigating screens).

CogAgent is a 9B-parameter bilingual vision-language GUI agent model based on GLM-4V-9B, trained with staged data curation, optimization, and strategy upgrades to improve perception, action prediction, and generalization across tasks. It focuses on operating real user interfaces from screenshots plus text, and follows a strict input–output format that returns structured actions, grounded operations, and optional sensitivity annotations. The model is designed for agent-style execution rather than freeform chat, maintaining a continuous execution history across steps while requiring a fresh session for each new task. Inference supports BF16 on NVIDIA GPUs, with optional INT8 and INT4 modes available but with noted performance loss at INT4; example CLIs and a web demo illustrate bounding-box outputs and operation categories.