Conversations: Love Them, Hate Them, Steer Them

TL;DR

This paper introduces a targeted activation engineering approach to steer large language models like LLaMA 3.1-8B towards more human-like emotional expression, enhancing their empathetic conversational abilities.

Contribution

It presents a novel method using attribution patching and emotional vectors to control emotional nuances in LLM responses without extensive fine-tuning.

Findings

Increased positive sentiment in responses

More frequent use of first-person pronouns

Enhanced emotional expressiveness

Abstract

Large Language Models (LLMs) demonstrate increasing conversational fluency, yet instilling them with nuanced, human-like emotional expression remains a significant challenge. Current alignment techniques often address surface-level output or require extensive fine-tuning. This paper demonstrates that targeted activation engineering can steer LLaMA 3.1-8B to exhibit more human-like emotional nuances. We first employ attribution patching to identify causally influential components, to find a key intervention locus by observing activation patterns during diagnostic conversational tasks. We then derive emotional expression vectors from the difference in the activations generated by contrastive text pairs (positive vs. negative examples of target emotions). Applying these vectors to new conversational prompts significantly enhances emotional characteristics: steered responses show increased positive sentiment (e.g., joy, trust) and more frequent first-person pronoun usage, indicative of greater personal engagement. Our findings offer a precise and interpretable method for controlling specific emotional attributes in LLMs, contributing to developing more aligned and empathetic conversational AI.