Astigmatism (Formerly 'Semantic Shift') - Astigmatism -0.6

Expanded the dataset for training by 50%, and used several training methods, using a preference based merge as the final lora.