Skip to content
Hot storyPast story

TaskFrame hierarchical planner-controller framework proposed

1 report1 sourceUpdated 23 d ago

What happened

AI overview

Guangzhi Wang et al. propose TaskFrame, a hierarchical planner-controller framework that connects task-level planning with frame-level keyboard and mouse control for game agents. A vision-language task planner breaks high-level goals into ordered sequences of atomic instructions and is optimized with supervised learning and dependency-aware GRPO; a causal Transformer controller outputs discrete key presses and continuous camera movement frame by frame. In Minecraft, its instruction decomposition is on par with general vision-language models, and its combat success rate in closed-loop evaluation is higher than comparison baselines.

AI-written from the coverage · updated 23 d ago

Coverage timeline

Follow the coverage to see the story from different sides.

Oct 29
  1. arXiv Game AI
    TaskFrame: Connecting task-level planning with frame-level control for game agents

    TaskFrame is a hierarchical planner-controller framework that connects task-level planning with frame-level keyboard and mouse control for game agents. A vision-language task planner breaks high-level goals into ordered sequences of atomic instructions and is optimized with supervised learning and dependency-aware GRPO; a causal Transformer controller outputs discrete key presses and continuous camera movement frame by frame. In Minecraft, its instruction decomposition is on par with general vision-language models, and its combat success rate in closed-loop evaluation is higher than comparison baselines.