Compositional Literary Primitives in Instruction-Tuned LLMs: Cross-Architectural SAE Features for Self, Style, and Affect

In three linesStudy of literary primitives in Llama 3.1 8B-Instruct and Gemma 2 9B-IT using sparse autoencoders. Four feature classes identified: naming-gates (affect tokens), self cluster (first-person register), stylistic modulators, compositional emotions. Llama achieves 27/27 emotion coverage (Cowen-Keltner taxonomy), Gemma 23/27. Validated via 5-LLM judge panel.

Summary generated by Claude — human-verified