AppWorld
A benchmark for evaluating LLM agents' ability to execute tasks in complex multi-application environments. It simulates real-world app operation settings and measures completion rates and efficiency across multi-step tool-calling workflows.
Timeline (last 90 days)
CoPE was validated on five benchmarks: ALFWorld, ScienceWorld, WebShop, τ²-bench, and AppWorld.
In the continuous self-improvement scenario of the AppWorld benchmark, JAZ's invoke outperforms ACE by 4% while also incurring lower costs.
AppWorld is an academic benchmark that simulates real-world application operating environments and is used to evaluate agents' continuous learning and self-improvement capabilities.
On the AppWorld benchmark, SMC reduces actual runtime by 7.7% compared to the SA baseline and by 44.9% compared to serial execution, with only a slight decrease in task completion rate.
All Facts (4)
CoPE was validated on five benchmarks: ALFWorld, ScienceWorld, WebShop, τ²-bench, and AppWorld.
50%UnverifiedIn the continuous self-improvement scenario of the AppWorld benchmark, JAZ's invoke outperforms ACE by 4% while also incurring lower costs.
50%UnverifiedAppWorld is an academic benchmark that simulates real-world application operating environments and is used to evaluate agents' continuous learning and self-improvement capabilities.
50%UnverifiedOn the AppWorld benchmark, SMC reduces actual runtime by 7.7% compared to the SA baseline and by 44.9% compared to serial execution, with only a slight decrease in task completion rate.
50%