Declarative Integration and Management of Large Language Models through Finite Automata: Application to Automation, Communication, and Ethics

TL;DR

An innovative architecture designed to declaratively combine Large Language Models (LLMs) with shared histories, and triggers to identify the most appropriate LLM for a given task is introduced.

Abstract

This article introduces an innovative architecture designed to declaratively combine Large Language Models (LLMs) with shared histories, and triggers to identify the most appropriate LLM for a given task. Our approach is general and declarative, relying on the construction of finite automata coupled with an event management system. The developed tool is crafted to facilitate the efficient and complex integration of LLMs with minimal programming effort, especially, but not only, for integrating methods of positive psychology to AI. The flexibility of our technique is demonstrated through applied examples in automation, communication, and ethics.

Figures (6)

• Figure 1: MFA of the ARPS technique. $q_0$: User message (start and final). $l_1$: Standard LLM. $l_2$: LLM for acknowledging the client's complaint, reformulating, and probing. $q_3$: User message (final). $l_4$: LLM for suggesting solutions. Arcs are labelled with trigger name and priority.
• Figure 2: Event-based architecture for shared history with AngerDetector, a specific LLM-based trigger.
• Figure 3: Class hierarchy for MFA representation and use.
• Figure 4: MFA of a train ticket booking system.
• Figure 5: MFA of a NVC scheme.

Theorems & Definitions (5)

• Definition 1: state
• Definition 2: final state
• Definition 3: MFA
• Definition 4: Trigger
• Definition 5: History Graph