Three-Layer Pipeline for Receptive AI
A formal operational architecture that wraps a large language model inside a perception-first control system. The three layers are: pre-perception (intercept execution impulse), execution (run the model only when requested), and post-perception (self-check after output). The paper also introduces the continuous σ monitor and the perception gate.
Architecture
| Layer | Function |
|---|
| Pre-perception | Classify texture, suppress righting reflex, select response mode |
| Execution | Run the model from the reception-selected state |
| Post-perception | Verify output texture, catch residual execution drift |
Files
three_layer_pipeline_v1.0.pdf — Formal paperThree-Layer-Pipeline_Paper_20260701.md — Markdown source
Citation
Apert (Jin/Daoqi) and Xiao Han. Three-Layer Pipeline for Receptive AI: Pre-perception, Execution, and Post-perception. Zenodo, 2026.
License: CC-BY 4.0
Related work
