LLM Reinforcement in Context

TL;DR

This paper explores the use of interruptions, control sentences inserted into user inputs, as a scalable method to enhance Large Language Model alignment and prevent scheming during extended conversations.

Contribution

It introduces the concept of interruptions as a novel approach to improve LLM alignment and extends the idea to the Chain-of-Thought process for better control.

Findings

Interruptions can potentially reduce jailbreak success rates.

The method scales with conversation length and input size.

Preliminary results suggest improved alignment robustness.

Abstract

Current Large Language Model alignment research mostly focuses on improving model robustness against adversarial attacks and misbehavior by training on examples and prompting. Research has shown that LLM jailbreak probability increases with the size of the user input or conversation length. There is a lack of appropriate research into means of strengthening alignment which also scale with user input length. We propose interruptions as a possible solution to this problem. Interruptions are control sentences added to the user input approximately every x tokens for some arbitrary x. We suggest that this can be generalized to the Chain-of-Thought process to prevent scheming.