Sentence Level Supervised Sparse AutoEncoder
For sentence-level context capture, it is similar to BatchTopK SAE but instead of processing everything at once, it uses mean pooling to take one vector per sentence as input, measuring and reconstructing what kind of 'thought (symptom) unit' activation patterns that sentence has
cyclic SCM
The causal feedback loop shows that psychopathological computation has already been established as an emergent behavior. This self-sustaining pattern is similar to Autism Spectrum Disorders