MPR-GUI-Bench reveals consistent P&R gaps between English and non-English GUI agent performance, and GUI-XLI reduces those gaps by an average of 6.5% via targeted hidden-state intervention during inference.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
MPR-GUI: Benchmarking and Enhancing Multilingual Perception and Reasoning in GUI Agents
MPR-GUI-Bench reveals consistent P&R gaps between English and non-English GUI agent performance, and GUI-XLI reduces those gaps by an average of 6.5% via targeted hidden-state intervention during inference.