Mind Lab puts continual learning to the test with Macaron-V1

The AI company said its model surpassed GLM-5.2 by training four billion additional parameters through specialized LoRA adapters.