What question did this study set out to answer?

The aim is to introduce an innovative AI architecture that supports transparent reasoning and learnability.

June 1, 20260 citationsOpen Access

NoLLM (Not‑only‑LLM) version 1 for SUNIE: Concept and Specifications of a White‑Box Next‑Generation AI

Key Points

The aim is to introduce an innovative AI architecture that supports transparent reasoning and learnability.
Introduced a structured mechanism comprising Concept Buckets and Function Chains.
Integrated classical rule-based methods with probabilistic rules to enhance stability in decision-making.
Outlined specifications for a hybrid AI system named SUNIE.
NoLLM architecture facilitates controllable and explainable AI behavior.
The integration of rule-based and probabilistic methods results in enhanced interpretability.
Conceptual foundations allow for effective auditing and observability throughout AI processes.

Abstract

This document presents “NoLLM (Not‑only‑LLM) version 1,” a white‑box next‑generation AI architecture designed for controllability, learnability, observability, and auditability. Unlike conventional LLM‑only systems, NoLLM introduces a structured mechanism composed of Concept Buckets, Function Chains, the Lottery mechanism, and the Kernel, enabling transparent and explainable reasoning processes. This specification outlines the conceptual foundations, internal representations, and system architecture of SUNIE, a hybrid AI system that integrates classical rule‑based methods with probabilistic rules to achieve stable and interpretable behavior.

Ask AI

Helpful

Bookmark

View Full Paper