APPSim-Bench: Bridging Real-world Apps and Reproducible Evaluation for Mobile GUI Agents
AppSim-Bench is proposed, which addresses this trade-off through controllable simulated apps that preserve task-relevant interaction logic while supporting deterministic evaluation and removes major sources of environmental stochasticity, enabling reproducible cross-model comparison.
Jin-Tian Feng, Long Chen, Xiao Yu et al.
· 0 citations