Retcon -- a Prompt-Based Technique for Precise Control of LLMs in Conversations

TL;DR

Retcon is a prompt-based method that enables precise, turn-level control of large language models during multi-turn conversations, improving their adaptability and performance over traditional prompting methods.

Contribution

The paper introduces Retcon, a novel few-shot prompting technique that allows dynamic control of LLM behavior in ongoing conversations, addressing a key challenge in conversational AI.

Findings

Retcon outperforms zero-shot prompting in controlling LLMs.

Retcon achieves significantly better results than traditional few-shot prompting.

Retcon enhances the flexibility and accuracy of LLMs in multi-turn interactions.

Abstract

Recent advances in Large Language Models (LLMs) allow agents to execute complex natural language tasks. Many LLM applications, such as support agents, teaching assistants, and interactive bots, involve multi-turn conversations. However, it remains challenging to control LLMs in the context of such interactions, particularly when the LLM behavior needs to be adjustable over the course of the conversation. In this paper, we present Retcon, a few-shot prompting technique designed to provide turn-level control over LLMs in conversations. We then demonstrate that it performs significantly better than zero-shot and traditional few-shot prompting.