TAG-Nav: A task-agnostic framework for mobile GUI navigation with graph memory
In brief
- TAG-Nav leverages graph memory to guide agents to the right pages before any task-specific instruction is given.
- DestBench evaluates navigation performance independently of downstream actions or task completion.
- Adding graph memory significantly improves navigation success rates for multiple models.
Mobile UI agents can perceive screens and execute actions but lack persistent navigation knowledge across apps. TAG-Nav introduces a task-agnostic graph navigation framework that uses destination memory to guide agents to the right pages before any specific task. DestBench, a new protocol, evaluates only the ability to reach the correct target page, isolating navigation performance.
What it changes for you
- Public SectorUI agents could automate navigation in administrative applications, reducing errors and speeding up processes.
- Human ResourcesHR management tools could benefit from agents capable of efficiently navigating complex HR interfaces.
This summary gives you the essentials. The full article, its figures and quotes are on cs.AI updates on arXiv.org.
Read the original articleAI-generated summary of the source, checked by rules. Report an error