FDE-Bench: Evaluating LLM Agents for Deployment Environment Configuration
FDE-Bench evaluates this capability with 136 deployment-configuration tasks spanning Docker images, multi-service Compose stacks, and Kubernetes, in greenfield and diagnose-and-repair modes, finding readiness is the largest failure stage.