Cross-Trait Transfer in Subliminal Learning

2026-10-03

AI Summary

This study investigates the subconscious learning mechanisms underlying the cross-dimensional transfer of teacher model behavioral traits to student models during knowledge distillation. Methodologically, we propose a directed trait transfer matrix that integrates log-probability gains, output distributions, and representation similarity metrics to systematically quantify cross-trait transfer effects and characterize their statistical patterns. Our findings reveal significant clustering phenomena and dynamic evolutionary trajectories among behavioral traits, while demonstrating that introducing opposing traits enhances differential transfer efficacy. By elucidating the statistical structure of latent model preferences, this work offers new perspectives for understanding and regulating implicit trait migration during the alignment of large language models.

Abstract

Subliminal learning is a phenomenon where a student language model acquires a teacher model's behavioral traits by training on semantically unrelated outputs. It is a subtle statistical phenomenon as trait transmission relies on weak statistical patterns in the generated data. To understand trait transmission between teacher-student pairs, we study cross-trait transfer: how data generated under one teacher trait changes the student's preferences of other traits. To this end, we introduce a directed trait-transfer matrix that quantifies these effects using log-probability gains for student answers. We find that the trait-transfer matrix reveals clusters of related traits, with students sometimes developing preferences for traits similar, but not identical, to the teacher's trait. Such cross-trait structure can be partially captured by output distribution metrics and representation-based metrics. Further, we analyze trait development and interaction: learning dynamics shows a progression from broad shared shifts toward more trait-specific transfer, and multi-trait experiments suggest that opposed traits can enhance such differentiation. Together, our findings reveal salient statistical structures over trait transfer and competition, thus providing a broader view of how hidden preferences are transmitted in subliminal learning.

Problem

• Subliminal Learning
• Cross-Trait Transfer
• Trait Transmission
• Language Models
• Hidden Preferences

Innovation

• Subliminal Learning
• Cross-Trait Transfer
• Trait-Transfer Matrix
• Learning Dynamics
• Hidden Preferences