Safety
Cross-Lingual Transfer of Evil Vectors Is Directionally Asymmetric
arXiv:2610.17880v1 Announce Type: cross Abstract: Persona vectors are linear directions in the residual-stream activations of LLMs that encode behavioral traits and enable fine-grained control over mo
arXiv:2610.17880v1 Announce Type: cross Abstract: Persona vectors are linear directions in the residual-stream activations of LLMs that encode behavioral traits and enable fine-grained control over model behavior. Among them, the evil persona vector is particularly relevant to understanding harmful behavior, jailbreak vulnerabilities, and the unlearning of harmful knowledge. Yet whether this vector reflects a genuinely cross-lingual concept, rather than an artifact of its extraction language, remains untested. We conduct a systematic evil-vector transfer study across two high-resource languages (English, Spanish) and two low-resource languages (Swahili, Telugu) on four open-source instruction-tuned LLMs, and find that transfer is directionally asymmetric. Evil vectors are harder to obtain reliably from low-resource languages, and the resulting vectors are largely ineffective at steering, even within their own language. In contrast, vectors extracted from high-resource languages transfer effectively to low-resource inputs while largely preserving general knowledge, as measured by MMLU. Geometric analysis further supports this pattern: evil vectors from the two high-resource languages are closely aligned, whereas low-resource vectors align only weakly with this shared direction. These results suggest that a shared cross-lingual evil direction exists and, once reliably extracted from a high-resource language, can be reused to steer behavior in low-resource languages without additional per-language data collection.
Related
- Emergent and Subliminal Misalignment Through the Lens of Data-Mediated Transfer
- FineSteer: A Unified Framework for Fine-Grained Inference-Time Steering in Large Language Models
- LLMs Encode Their Failures: Predicting Success from Pre-Generation Activations
- Evil Spectra: How Optimisers can Amplify or Suppress Emergent Misalignment
- Your LLM, Your Style: Behavioral Mode Axes for LLM Behavioral Control
Source: arXiv cs.AI | 2026-10-15