Synthetic Intuition: A System-1/System-2 Architecture for Fast and Slow Thinking in Large Language Models

Current large language models (LLMs) apply uniform computational resources to every token generation, regardless of task complexity-a fundamental inefficiency that contradicts human cognitive architecture. Inspired by Kahneman's dual-process theory, we propose Synthetic Intuition, a novel architecture that integrates a fast, intuitive System1 model with a deliberate, reasoning-oriented System-2 model. System-1, a lightweight transformer (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{3 5 0 M}$</tex> parameters), handles routine token generation with high confidence, while System2 (7B parameters) is selectively activated when System-1 detects uncertainty or complex reasoning requirements. Our architecture achieves a <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$3.2 \times$</tex> speedup in inference while maintaining 97.3 % of the original model's performance on diverse benchmarks. On reasoning-intensive tasks (GSM8K, ARC-Challenge), our approach matches or exceeds baseline performance with 68 % reduced computational cost. We demonstrate that cognitive load can be dynamically allocated, opening new pathways for efficient and adaptive language model deployment.