Researchers release SkillChain-Gym, a benchmark for reskilling-aware production-inventory control under disruptions

Researchers have introduced SkillChain-Gym, a benchmark specification for reskilling-aware production-inventory control under disruptions. The benchmark models workforce skill-state dynamics, including hard-threshold certifications, forgetting effects, and capacity-consuming training actions constrained by the same per-worker time budget as production. It includes seed-controlled disruption scenarios and three feasibility modes with projection diagnostics, as well as deterministic replay and metrics spanning operations, resilience, capability growth, and training-access distribution.

The benchmark evaluates four policy classes—production-only, reactive adaptive, water-filling adaptive, and static-insurance policies—with budget variants over 60-shift horizons using paired statistical tests. Results show training-capable policies dominate production-only baselines, and maintenance training is necessary even without disruptions due to forgetting. Among training-capable classes, adaptive training performs best when forecasted bottlenecks are visible, while a lean static cross-training plan serves as strong insurance under surprise shocks and absenteeism.

Capacity slack and the forgetting rate determine the boundary between regimes where different policy classes excel. No single policy class dominates across all regimes, motivating the development of forecast-driven controllers that dynamically decide when to invest in skill maintenance versus reactive adaptation.

The benchmark is designed as a reusable testbed for operations research, addressing a gap where workforce-planning models with skills and learning are rarely released as standardized environments. It provides a controlled setting to study the trade-offs between production throughput, workforce reskilling, and resilience under varying disruption scenarios.