QUALS: Corpus Equilibrium for Universal Forecasting via Pattern Quantization and Learnability Synchronization
QUALS improves zero-shot time-series foundation model training via pattern quantization and learnability synchronization, achieving better accuracy with a fraction of the data.