ToolTok: Tool Tokenization for Efficient and Generalizable GUI Agents

Xiaoce Wang, Guibin Zhang, Junzhe Li, Jinzhe Tu, Chun Li, Ming Li

🏛 Institutions: Tsinghua, NUS, PKU, Shenzhen MSU-BIT University, Guangming Laboratory
📅 Date: January 30, 2026
📑 Publisher: arXiv
💻 Env: General GUI
🔑 Keywords: tool use tokenization curriculum learning visual grounding data efficiency ToolTok

TLDR

ToolTok treats GUI operations as paths over learnable tool tokens with semantic anchoring and curriculum learning. This makes GUI agents more efficient and generalizable, reaching competitive performance with far less training data than other post-training methods.

Open paper arXiv Edit on GitHub Report issue