Summary of Autodroid-v2: Boosting Slm-based Gui Agents Via Code Generation, by Hao Wen et al.
AutoDroid-V2: Boosting SLM-based GUI Agents via Code Generation
by Hao Wen, Shizuo Tian, Borislav Pavlov, Wenjie Du, Yixuan Li, Ge Chang, Shanhui Zhao, Jiacheng Liu, Yunxin Liu, Ya-Qin Zhang, Yuanchun Li
First submitted to arxiv on: 24 Dec 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary In this paper, researchers propose a new approach to building mobile UI agents that can complete arbitrary natural language tasks through interactions on users’ devices. The goal is to reduce the need for powerful large models by customizing smaller domain-specific models using high-quality training data. The team draws inspiration from recent small language models and converts the UI task automation problem into a code generation problem, which can be solved on-device. To overcome the challenges of generating UI automation code, they adopt a document-centered approach that automatically builds API documentation for each app and generates diverse task samples based on this documentation. By guiding the agent with synthetic documents and task samples, it learns to generate precise and efficient scripts to complete unseen tasks. The proposed approach outperforms state-of-the-art mobile UI agents in terms of success rates and latency/token consumption. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Mobile UI agents can now automate arbitrary natural language tasks on devices without needing powerful large models. This is achieved by customizing smaller domain-specific models with high-quality training data. By converting the task automation problem into a code generation problem, an on-device small language model (SLM) can solve it efficiently. To generate UI automation code, the team uses API documentation for each app and generates diverse task samples based on this documentation. This approach guides the agent to create precise scripts that complete tasks without needing to be trained extensively with public datasets. |
Keywords
» Artificial intelligence » Language model » Token
