LLM Layer Classifier An experimental framework for exploring whether there is subspace separability between layers of embeddings in LLM. post see: lesswrong You can find demo pipeline in 0_unittest.ipynb