LOADBench
An open-source dataset and standardized benchmark for analog IC design, built on the SkyWater130 PDK. It covers over a hundred operational amplifier topologies and defines six evaluation tasks to support fair comparison of machine learning methods.
Timeline (last 90 days)
LOADBench provides a large-scale open-source analog circuit dataset built on the SkyWater130 PDK, containing over 100 operational amplifier topologies, covering single-stage, two-stage, and three-stage structures.
Each op-amp topology in the LOADBench dataset comes with structural annotations, functional annotations, topology-specific sizing constraints, and a directly runnable Ngspice testbench.
Each topology in the LOADBench dataset comes with the results of one multi-objective NSGA-II sizing optimization run, generating about 10³ Pareto-optimal solutions and 10⁶ intermediate sizing-performance pairs.
LOADBench defines six standardized benchmark tasks: inverse sizing design, topology selection, subcircuit classification, structured circuit generation, Pareto membership prediction, and multi-objective optimization.
LOADBench's six benchmark tasks all come with clear input/output definitions and evaluation metrics, and the reference implementation is released under an open-source license.
The evaluation metrics for LOADBench's inverse sizing design task typically include the final performance compliance rate, the simulation verification pass rate, and the closeness to the true Pareto front.
LOADBench's dataset, annotated ground truth, and the reference implementation of its metric definition routines are all released under an open license.
All Facts (7)
LOADBench provides a large-scale open-source analog circuit dataset built on the SkyWater130 PDK, containing over 100 operational amplifier topologies, covering single-stage, two-stage, and three-stage structures.
50%UnverifiedEach op-amp topology in the LOADBench dataset comes with structural annotations, functional annotations, topology-specific sizing constraints, and a directly runnable Ngspice testbench.
50%UnverifiedEach topology in the LOADBench dataset comes with the results of one multi-objective NSGA-II sizing optimization run, generating about 10³ Pareto-optimal solutions and 10⁶ intermediate sizing-performance pairs.
50%UnverifiedLOADBench defines six standardized benchmark tasks: inverse sizing design, topology selection, subcircuit classification, structured circuit generation, Pareto membership prediction, and multi-objective optimization.
50%UnverifiedLOADBench's six benchmark tasks all come with clear input/output definitions and evaluation metrics, and the reference implementation is released under an open-source license.
50%UnverifiedThe evaluation metrics for LOADBench's inverse sizing design task typically include the final performance compliance rate, the simulation verification pass rate, and the closeness to the true Pareto front.
50%UnverifiedLOADBench's dataset, annotated ground truth, and the reference implementation of its metric definition routines are all released under an open license.
50%