Search papers, labs, and topics across Lattice.
The authors evaluate the feasibility of training-free, additive steering vector composition across language, jailbreaking, and conciseness across four instruction-tuned LLMs from two model families and scales. They discover that successful multi-attribute composition requires layer-stratified intervention, where linguistic control operates best in early layers while abstract behavioral traits peak in middle layers. Geometric analysis reveals that these vectors maintain approximate mutual orthogonality in the residual stream, allowing simultaneous multi-attribute steering to scale up to three concurrent attributes without catastrophic interference.
Steering vectors for distinct model behaviors remain naturally orthogonal in the residual stream, enabling training-free, multi-attribute control over language, safety, and style simply by stratifying injection across model depth.
This study examines the compositionality of steering vectors for language and behavioral control in large language models. Focusing on language, jailbreak, and conciseness, we investigate whether additive, training-free composition of attribute steering vectors can preserve the intended steering effect of each attribute, across four instruction-tuned models from two model families and two size scales. We find that single-attribute steering is reliable for all three attributes, but only within an appropriate combination of intervention layer and steering strength, with abstract behaviors (jailbreak, conciseness) favoring middle layers and language favoring earlier layers. We show that additive composition of two attribute vectors succeeds in steering both attributes simultaneously when each is injected at its own best-performing layer, and that this partially extends to three simultaneously composed attributes, addressing an inconsistency left open by prior work on training-free composition. We further analyze the geometric properties of these steering vectors, finding that they are approximately orthogonal in the residual stream, consistent with their compositional behavior.
