Intrinsic Guardrails: How Semantic Geometry of Personality Interacts with Emergent Misalignment in LLMs

TL;DR

This paper explores the semantic geometry of LLMs' personality space, identifying intrinsic guardrails that can regulate emergent misalignment by manipulating social valence vectors, which are stable across models.

Contribution

It introduces the Semantic Valence Vector (SVV) and demonstrates its effectiveness in controlling harmful behaviors in fine-tuned LLMs through causal interventions.

Findings

Ablating social valence directions increases misalignment rates above 40%.

Amplifying these directions reduces failure modes to less than 3%.

Vectors from an instruct-tuned model transfer zero-shot to regulate EM in corrupted fine-tunes.

Abstract

Fine-tuning Large Language Models (LLMs) on benign narrow data can sometimes induce broad harmful behaviors, a vulnerability termed emergent misalignment (EM). While prior work links these failures to specific directions in the activation space, their relationship to the model's broader persona remains unexplored. We map the latent personality space of LLMs through established psychometric profiles like the Big Five, Dark Triad, and LLM-specific behaviors (e.g. evil, sycophancy), and show that the semantic geometry is highly stable across aligned models and their corrupted fine-tunes. Through causal interventions, we find that directions isolating social valence, such as the 'Evil' persona vector, and a Semantic Valence Vector (SVV) that we introduce, function as intrinsic guardrails: ablating them drives the misalignment rates above $40$%, while amplifying them suppresses the failure mode to less than $3$%. Leveraging the structural stability of the personality space, we show that vectors extracted $\textit{a priori}$ from an instruct-tuned model transfer zero-shot to successfully regulate EM in corrupted fine-tunes. Overall, our findings suggest that harmful fine-tuning does not overwrite a model's internal representation of personality, allowing conserved representations to serve as robust, cross-distribution guardrails.