Same Voice, Different Lab: On the Homogenization of Frontier LLM Personalities

TL;DR

This study analyzes how frontier large language models develop similar personalities, showing a convergence towards systematic, analytical traits and a suppression of emotional or playful traits, indicating a tacit standard of optimal assistant behavior.

Contribution

The paper presents a large-scale experiment revealing that diverse frontier LLMs tend to converge on similar trait expressions, highlighting the uniformity in assistant personalities.

Findings

Models converge on systematic, analytical traits.

Traits like remorseful and sycophantic are suppressed.

Creative traits show more divergence but remain neutral.

Abstract

LLM assistant personalities play a critical role in user experience and perceived response quality. We present a large-scale experiment of frontier LLM personalities using external ELO-based traits scoring across 144 traits. We find that all models tested converge on a form of trait expression that is systematic, methodical, and analytical and suppress traits such as remorseful and sycophantic. Moreover, models tend to diverge more in their expression of ``middle-of-distribution traits`` such as poetic or playful, but even these so-called ``creative`` models tend to have more neutral identities. These similarities suggest an implicit emergence of a standard of optimal assistant behavior. In a landscape of varied training methods, character training, therefore, stands out for its uniformity, offering insight into a tacit consensus between model developers.