SiliQunLib: Pre-Trained Primitive Gate Policy Checkpoints for Silicon Spin Qubit Control

SiliQunLib is a curated repository of 50 pre-trained Soft Actor-Critic (SAC) actor-network checkpoints for primitive quantum gate synthesis on silicon spin qubit hardware. Each checkpoint encodes a policy trained to synthesise a specific entanglement-family target state (Bell, GHZ, W, Cluster, Dicke-k2, Dicke-k3) across the 2–5 qubit range, using the SiMOS-nominal hardware profile from the SiliQun simulator. These checkpoints are the companion artefact for the paper: R. M. Alshehri, 'SiliQun: A Silicon Spin Qubit Simulator and Reinforcement Learning Environment for Quantum Gate Synthesis,' Quantum Science and Technology, 2026 (under review). Version 1 contains architecturally correct placeholder weights (Xavier-uniform initialisation). Trained weights will be provided in v2 following completion of the full training sweep on Aziz HPC.